Datasets:

Galahad3x

/

QADatasetForPatho

like 0

200

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

How long did the event last?

201

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

In which street did this happen?

202

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

In which city did this happen?

203

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

In which region did this happen?

204

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

In which country did this happen?

205

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

Where did this happen?

206

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

What caused the event?

207

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

What was the cause of the event?

208

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

What source started the event?

209

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

How was the event first detected?

210

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

How many people were ill?

211

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

How many people were hospitalized?

212

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

How many people were dead?

213

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

Which contaminants or viruses or bacteria were found?

214

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

Which were the symptoms?

215

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

Which were the symptoms?

216

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

What did the patients have?

217

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

What were the first steps?

218

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

What did they do to control the problem?

219

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

What did the local authorities advise?

220

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

What were the control measures?

221

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

What type of samples were examined?

222

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

What did they test for in the samples?

223

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

What is the concentration of the pathogens?

224

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

What steps were taken to restore the problem?

225

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

What was done to fix the problem?

226

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

What could have been done to prevent the event?

227

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

How to prevent this?

228

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

What were the investigation steps?

229

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

What did the investigation find?

230

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

How was the infrastructure affected?

231

A large and persistent outbreak of typhoid fever caused by consuming contaminated water and street-vended beverages Kampala, Uganda, January

Abstract Background: On 6 February 2015, Kampala city authorities alerted the Ugandan Ministry of Health of a “strange disease†that killed one person and sickened dozens. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods: We defined a suspected case as onset of fever (≥37.5 °C) for more than 3 days with abdominal pain, headache, negative malaria test or failed anti-malaria treatment, and at least 2 of the following: diarrhea, nausea or vomiting, constipation, fatigue. A probable case was defined as a suspected case with a positive TUBEX® TF test. A confirmed case had blood culture yielding Salmonella Typhi. We conducted a case-control study to compare exposures of 33 suspected case-patients and 78 controls, and tested water and juice samples. Results: From 17 February-12 June, we identified 10,230 suspected, 1038 probable, and 51 confirmed cases. Approximately 22.58% (7/31) of case-patients and 2.56% (2/78) of controls drank water sold in small plastic bags (ORM-H = 8.90; 95%CI = 1.60–49.00); 54.54% (18/33) of case-patients and 19.23% (15/78) of controls consumed locallymade drinks (ORM-H = 4.60; 95%CI: 1.90–11.00). All isolates were susceptible to ciprofloxacin and ceftriaxone. Water and juice samples exhibited evidence of fecal contamination. Conclusion: Contaminated water and street-vended beverages were likely vehicles of this outbreak. At our recommendation authorities closed unsafe water sources and supplied safe water to affected areas. Keywords: Typhoid fever, Outbreak, Case-control, Uganda Background Typhoid fever is a systemic disease caused by Salmonella enterica serovar Typhi, a Gram-negative bacterium. Humans are the only host, and transmission most commonly occurs through ingestion of water or food contaminated by feces from an acutely ill or convalescent patient or an asymptomatic carrier. The incubation period is usually 1 to 2 weeks but can range from 3 to 60 days [1]. The illness presents with sustained fever and a constellation of other symptoms including dry cough, fatigue, abdominal pain, diarrhea, and constipation [2]. Case fatality ratios range between 10 and 30% if untreated, but fall to 1–4% with appropriate and timely antimicrobial treatment [3]. The gold standard laboratory diagnosis of typhoid fever requires isolation of S.. Typhi from blood, stool, bone marrow, or other tissue or bodily fluid by bacterial culture [2]. Other tests with moderate sensitivity and specificity include the Widal test and TUBEX® TF test which involve detection of antibodies against S. Typhi antigens [2]. Typhoid fever is preventable through public health interventions such as provision of safe water, ensuring proper sanitation and waste disposal systems, and excluding disease carriers from handling food [4]. Typhoid fever is a major cause of mortality and morbidity worldwide. In endemic areas, the disease is most commonly found in children 5–19 years of age. International visitors from non-endemic areas are also at risk if unvaccinated [1]. The global burden of the disease in low- and middle-income countries in 2010 was estimated to be 11.9 million cases, including 129,000 fatalities, after adjusting for water-related risk factors [5]. In Uganda, an outbreak of typhoid fever in Kasese District sickened 8092 persons from 27 December 2007 to 30 July 2009, resulting in at least 249 intestinal perforations and 47 deaths [6]. In 2011, numerous typhoid cases were again reported in Kasese and neighboring Bundibugyo District with many more intestinal perforations and emergence of multidrug resistant strains [7]. On 6 February, 2015, the Ugandan Ministry of Health (MoH) received a report from the Kampala Capital City Authority that a 42-year-old man had died a day earlier of a “strange illness.†The patient was admitted to the hospital on 2 February 2015 with symptoms of abdominal pain, high fever, and severe jaundice. Initial testing involved use of the Widal test which was positive. Approximately 30 other people who worked in the same area as the deceased reportedly had similar symptoms. We conducted an epidemiologic investigation to identify the nature of the disease, mode of transmission, and risk factors to inform timely and effective control measures. Methods Study sites The outbreak occurred in Kampala (estimated population 1.4 million), the capital of Uganda [8]. Kampala has five divisions: Kampala Central, Kawempe, Makindye, Rubaga, and Nakawa. We focused our epidemiologic investigation on two markets and a commuter taxi park in Kampala Central Division where the initial cases were concentrated. Surveillance To characterize and control the epidemic, MoH conducted surveillance at six treatment centers established in affected areas of the city to provide diagnostic testing and typhoid fever treatment free of charge. These treatment centers were existing health centers in which routine disease surveillance and treatment activities are conducted. Through the media, local leaders encouraged the people with symptoms of typhoid fever to seek medical care at these treatment centers. We defined a suspected case as onset of fever (≥37.5 °C) for ≥3 days from 1 January 2015 onwards, with headache, abdominal pain, a negative test for malaria or failure to respond to anti-malaria treatment, and ≥2 of the following symptoms: diarrhea, nausea or vomiting, constipation, or fatigue. A probable case was a suspected case whose serum sample yielded a positive TUBEX® TF test [9]. Blood samples were collected from the first 5 suspected cases every day from each treatment center and referred to the microbiology laboratory at the Medical Research Council for blood culture. A confirmed case was a suspected case whose blood culture yielded S. Typhi. Case-control study We conducted open-ended hypothesis-generating interviews of case-patients found in the areas where the initial cases were identified, focusing on their usual sources of water and food. To test the hypotheses generated from the interviews, we conducted a case-control study from 10 to 20 February 2015. To rapidly identify the mode of transmission so as to inform prompt prevention and control measures, we used the initial 33 suspected case-patients identified in the earliest-affected communities for the case-control study. The earliest cases were persons working in two markets or in the central terminal station for Kampala’s shared taxis, all of which were located in central Kampala. Therefore we recruited both the cases and the controls from those places. The markets are open spaces where people set up their stalls to sell assorted merchandise, whereas the central terminal station for the shared taxis is an area where the shared taxis (mini-vans) pick up and drop off passengers. In the markets, after identifying and interviewing a case, the interviewer then walked around the stall to identify several persons of the same gender and similar age as the case from the surrounding stalls who never had a febrile disease since January 1, 2015, and recruited those persons as controls. Similarly, in the central terminal station for the shared taxis, after identifying and interviewing a case who was working inside a shared taxi (e.g., a driver or conductor), the interviewer then walked around the shared taxi to recruit asymptomatic workers of the same gender and similar age from the surrounding shared taxis as controls. The interviewers used a structured questionnaire to collect information on the usual water and food exposures from the case- and control-persons. A link to the questionnaire that was used has been provided in the Additional file 1 of the manuscript. Clinical laboratory investigation The TUBEX® TF test was performed at the treatment centers by trained clinical and laboratory staff as per the manufacturer’s instructions. Blood culture was performed on the first five patients presenting each day at the 6 treatment centers. From each adult patient, 5–10 mL of blood was collected and inoculated in BD Bactecâ„¢ Aerobic/F blood culture bottles and incubated in a BD Bactec 9000 seriesâ„¢. Presumptive positive bottles, as signaled by the system, were subcultured on MacConkey, chocolate, and blood agar plates and incubated aerobically at 37 °C for 24 h. A Gram stain was also performed. Negative vials were incubated for up to 7 days and if the system still indicated negative, a Gram stain was performed and a final subculture was done before reporting the specimen as negative. Oxidase-negative, lactose non-fermenting colonies, were screened using API 10S at the start of the outbreak. Later an abbreviated panel of biochemical tests [10] was used. Isolates biochemically typical of S. Typhi were serotyped using slide agglutination with S. polyvalent O, S. polyvalent H, S. O factor 9 (group D), S. H factor d and S. Vi antisera. A set of 30 S. Typhi isolates were sent to the U.S. Centers for Disease Control and Prevention (CDC) for confirmation and antimicrobial susceptibility testing (AST). The National Antimicrobial Resistance Monitoring System at CDC performed AST on 17 isolates by broth microdilution to determine minimum inhibitory concentrations for 14 antimicrobial agents: amoxicillin/clavulanic acid, ampicillin, azithromycin, ceftiofur, ceftriaxone, cefoxitin, chloramphenicol, ciprofloxacin, gentamicin, nalidixic acid, streptomycin, sulfisoxazole, tetracycline, and trimethoprim/sulfamethoxazole. Results were interpreted using Clinical and Laboratory Standards Institute standards [11] when available. During the case-control study, we collected 5– 10 mL of blood from each of 20 suspected casepatients, placed the samples into commercial BD Bactecâ„¢ Aerobic/F media, and transported them to the clinical laboratory at the Makerere College of Health Sciences Department of Medical Microbiology for incubation in the Bactec 9120â„¢ blood culture system. Subcultures onto MacConkey and blood agar were done following instrument signals of growth or at the end of 7 days of incubation. Colonies were identified as S. Typhi based on biochemical characteristics including motility, hydrogen sulfide production, fermentation of sugars, urease production, and serological typing characteristics with various specific antisera. Environmental laboratory investigation From 2 to 8 April, juice samples were collected from the Nakasero, Owino, and Shauriyako markets, and 100 mL water samples were collected from unprotected water sources such as unprotected springs (i.e., underground water sources that do not have barriers protecting them from contamination and run-off) and commercial vendors in Kampala Central Division. We chose these water collection sites because we observed people in the outbreak-affected areas collecting water from these sites. The juice samples were tested because case-persons said they usually consumed these drinks. We collected nine juice samples, including 3 “bushera†(millet and yeast), 2 “munanansi†(pineapple juice with tea leaves), 3 “butunda†(passion fruit), and 1 “bongo†(unpasteurized yogurt drink). We also collected 13 water samples, including 3 “kaveera†(water packaged and sold in a small plastic bag), one unlabeled bottle of water from a street vendor, water from three storage tanks, and water from five unprotected springs. Juice and water samples were tested using a modified version of the United States Environmental Protection Agency’s Standard Analytical Protocol for S. Typhi in Drinking Water [12]. Briefly, 125 mL of specimen was preenriched in 125 mL of double strength buffered peptone water at 37 °C, followed by parallel enrichment in Selenite Cysteine broth at 37 °C and RV broth at 42 °C. Cultures from Selenite Cysteine broth were plated onto MacConkey and XLD agars; cultures from RV broth were plated onto XLD agar. All plates were incubated at 37 °C. Plates were inspected at 24 and 48 h for colony morphology consistent with Enterobacteriacea. Colonies morphologically consistent with S. spp. (i.e. lactose negative) were subjected to biochemical testing. Suspect isolates were sent to CDC-Atlanta for biochemical confirmation. For confirmation, suspect cultures were streaked onto Hektoen enteric agar and suspect colonies were subjected to an abbreviated panel of tests, for phenotypic identification of Salmonella or Shigella spp. and biochemical differentiation of S. serovars Typhi and Paratyphi A from other Salmonella serovars [10]. Statistical analysis Using surveillance data, the attack rates by sex, division, and sex were calculated using population data from the national census [8] and data provided by the Uganda Bureau of Statistics [13]. Using the StatCalc in Epi Info 7, considering a power of 80%, two sided confidence level of 95%, a case-control ratio of 1:2 with 30% of cases exposed and 10% of controls exposed, we would require about 39 cases and 77 controls. To measure the associations between exposure variables and illness status, we used the Mantel-Haenszel method to estimate odds ratios (OR) and their confidence intervals, accounting for frequency-matching of cases and controls. We calculated the proportion of cases and controls who drank 1, 2, and 3–4 types of locally made drinks, and used the Chisquare test for linear trend to assess the relationship between the number of types of drinks consumed and odds of illness [14]. Results Surveillance From 17 February to 12 June 2015, we identified 10,230 suspected cases from the six treatment centers established by MoH. Cases were distributed widely throughout Kampala and neighboring areas (Fig. 1). The epidemic curve of suspected cases suggests that the outbreak started at the beginning of February or perhaps earlier. By the time the outbreak was recognized on 6 February, hundreds of cases had already occurred (Fig. 2). Cases were reported in all five divisions of Kampala: Makindye (32%, 3234), Rubaga (28%, 2828), Kawempe (11%, 1144), Nakawa (6.4%, 656) and Central (4.2%, 428); for 19% (1940) of cases, either no division of residence was identified, or resided outside of Kampala. The attack rate during the outbreak period was highest in Makindye (10/1,000), Rubaga (8.7/1000), and Central (6.5/1000) Divisions. Males had a higher attack rate than females. The attack rate among people in the 15–59 year age group (12/1000) was 6 times higher than among younger (2.0/1000) or older (2.0/1000) persons (Table 1). Case-control study In our hypothesis-generating interviews of patients from the area where the outbreak was first identified, consumption of drinks made with water extracted from unprotected sources and packed in unhygienic conditions was often reported. Of the 33 case-patients we enrolled in the case-control study, 60% were men; the majority of the case-patients (85%) were in the 20–39 year and 9.1% in older age groups. In addition to fever, commonly reported symptoms included abdominal pain (72.72%) and headache (69.69%) (Table 2). We found that 22.58% (7/31) of case patients compared with 2.56% (2/78) of controls usually drank locally packaged water in small plastic bags called “kaveera water†(ORM-H = 8.90; 95% CI = 1.60–49.00); 55% (18/33) of case-patients compared with 19.23% (15/78) of controls drank locally-made passion fruit juice called “butunda†(ORM-H = 4.60; 95% CI: 1.90– 11.00); 31.25% (10/32) of case-patients compared with 16.67% (13/78) of controls usually drank locally-packed pineapple juice called “munanansi†(ORM-H = 2.00; 95% CI = 0.74–5.20); and 15.63% (5/32) of case-patients compared with 8.97% (7/78) of controls usually drank cold millet porridge called “bushera†(ORM-H = 2.80; 95% CI = 0.76–10.00). Workplace as a source of breakfast (ORM-H = 0.25; 95% CI = 0.07–0.93), and workplace as a source of lunch (ORM-H = 0.35; 95% CI = 0.11–1.10) were not significant risk factors for illness. When we compared the proportions of case-patients and controls who drank 0, 1, 2, or 3–4 types of locallymade drinks, we found that case-patients were more likely to drink multiple types of locally-made drinks than controls (Chi-square for linear trend = 14.65, p < 0.001) (Table 3). Laboratory investigation Of 10,230 suspected cases, 3464 (10%) underwent TUBEX® TF testing. Of those, 1038 were positive, representing a positivity rate of 29%. Blood samples from a total of 364 patients (including 20 of 33 case-control study patients) were tested by blood culture and 56 (15%) (including 5 of the 20 case-control study patients tested) yielded S. enterica ser. Typhi. Subsequently, 30 of the 56 S. Typhi isolates from blood cultures were confirmed at the U.S. CDC as S. Typhi. CDC determined the minimum inhibitory concentrations for 17 of these isolates, 5 of which were resistant to ampicillin, chloramphenicol, streptomycin, sulfisoxazole, nalidixic acid, trimethoprim/sulfamethoxazole and had intermediate interpretation to ciprofloxacin. The remaining 12 were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. Environmental investigation One of 3 “kaveera water†samples and the unlabeled bottle of water sold by street vendors contained lactose fermenting bacteria, which are commonly Enterobacteriaceae and associated with fecal contamination. The 2 other “kaveera water†samples contained non-lactose fermenting bacteria, also consistent with fecal contamination, and one isolate was further identified as non-typhoidal Salmonella. The 5 water samples from unprotected springs showed evidence of robust contamination with lactose fermenting bacteria. Lactose non-fermenting colonies from 2 spring water samples were identified as non-typhoidal Salmonella spp. Lactose fermenting bacteria were also detected in 2 of 3 “bushera†samples, 1 of 2 “munanasi†samples, all 3 “munanansi†juice samples, and the “bongo†sample. Additionally, non-typhoidal Salmonella was cultured from 1 “busheraâ€, 1 “munanansiâ€, and 1 passion fruit juice sample. Discussion Our investigation revealed a prolonged and widespread outbreak of typhoid fever that affected thousands of people in all five divisions of Kampala City over several months. Contaminated water from unprotected sources and drinks made with it were the likely vehicles of infection early in the outbreak. Juice and water samples obtained from street vendors and water samples collected from unprotected spring water sources showed evidence of fecal contamination. Although S. Typhi was not recovered from environmental testing, non-Typhi Salmonella were isolated from five street-vended beverage samples from the implicated markets. All 17 isolates of S. Typhi from blood tested at CDC were resistant to nalidixic acid and had intermediate interpretation to ciprofloxacin. It is possible that persons affected by the antibiotic-resistant strains during this outbreak had experienced complications of typhoid fever considering that antibiotic-resistant strains of S. Typhi are associated with more severe form of the illness, complications and death [15]. This outbreak may have started in January 2015 or even earlier; however, it was not recognized until early February because routine clinical and laboratory surveillance systems for typhoid fever were not in place before the investigation. The widespread nature of the outbreak is compatible with a waterborne source. The sudden increase in cases after the start of the investigation was likely due to active community outreach and education about the symptoms of typhoid fever and the availability of prompt, free diagnostic testing and treatment through newly established treatment centers. The gradual decline in cases from mid-March onwards was likely the result of patient treatment and public health interventions including provision of free water chlorination products, sensitization of residents on water treatment, and the establishment of free alternative safe water sources in the most affected communities. Based on the evidence we presented, the Kampala Capital City Authority sealed off all underground water sources and worked with the National Water and Sewerage Corporation to ensure the provision of accessible alternative sources of water to the affected communities. In Uganda, as in many low and middle income countries, definitive diagnostic tests for typhoid fever such as blood culture are usually unavailable, unaffordable, or inconsistently applied [16]. Instead, typhoid fever diagnosis and surveillance often rely on clinical judgment or on the Widal test, which has poor sensitivity and specificity [17]. Moreover, physicians often give presumptive antibiotic and/or antimalarial treatment for febrile illnesses [18, 19] without attempting to determine the etiology. Previous studies have indicated that a significant proportion of febrile illness in Uganda is caused by bacteremia, including invasive non-Typhi salmonellosis and typhoid fever [20]. A more robust approach in these settings could entail periodically identifying persons with febrile illness in the communities and taking blood culture for confirmation [21]. The blood samples could be collected and sent using a specialized transport network to regional laboratory centers around the country where confirmative tests can be performed. This system has been successfully used to improve diagnostic services in early infant HIV/AIDS diagnosis [22]. Sentinel surveillance for febrile illnesses based on blood cultures would accelerate the early identification of outbreaks and implementation of control measures. Waterborne typhoid and paratyphoid fever affect an estimated 27 million people worldwide each year [3]. In developing countries, where safe water and sanitation systems have not been well-established, large-scale typhoid and paratyphoid outbreaks sometimes occur [7, 23–26]. During a previous typhoid outbreak in Kasese and Bundibugyo districts, Uganda, in 2009–2011, which affected 8092 persons, the vehicle of transmission was also found to be unclean water [7]. The current outbreak was likely caused by consuming contaminated water from unprotected ground water sources. Kampala city has more than 200 unprotected ground water sources, most of which serve as unprotected sources of water for economically disadvantaged people in the city such as those in our investigation [27]. Unsafe disposal of excreta and solid waste are significant factors that contribute to contamination of ground water in Kampala [28]. This outbreak investigation highlights the importance of ensuring access to affordable, safe, treated drinking water and improved sanitation and waste management systems for resource-constrained urban populations. Risk factors for typhoid transmission were not assessed later during this outbreak, when foodborne transmission might have become more common. Recurrent contamination of unprotected water sources with S. Typhi likely continued to sustain the outbreak propagation over the course of several months. According to the Uganda Demographic Health Survey 2011 [29], almost 30% of people living in urban areas and more than 60% of those living in rural areas do not treat their water before drinking it. Barriers to safer drinking water include the cost associated with establishing a piped treated water system or purchasing water treatment products for household use and the false perception that naturally occurring water sources could be safe [30]. In the aftermath of outbreaks like this one, public health authorities face 3 possible options: The first option is to do nothing but respond to outbreaks as they occur. Governments in resource-constrained settings often choose this option, leaving the population vulnerable to outbreaks of waterborne diseases including cholera, hepatitis A and E, cryptosporidiosis, shigellosis, and many others in addition to typhoid and paratyphoid fever. The second option is mass vaccination against typhoid fever. A cost-effectiveness evaluation of a hypothetical typhoid vaccination campaign was carried out after the multi-year outbreak of typhoid fever in Kasese District, Uganda, and it was estimated to be highly costeffective [31]. However, vaccination against the many different pathogens that cause waterborne diseases is not possible because vaccines are not yet available for many of them (e.g. cryptosporidiosis, shigellosis, paratyphoid fever, etc.). In addition, typhoid fever vaccines have been shown to have varied levels of effectiveness (from 50 to 95%) and to last for varied lengths of time (from 3 to 10 years) [15]. The third and final option is to improve the water and sanitation systems. Improvement of sanitation, hygiene and clean water supply around the world could avert ≥90% of diarrheal disease episodes annually [32]. In North America and Europe, typhoid fever caused largescale outbreaks from the late 19th through the early 20th century [33, 34]. After improvement of municipal water and sanitation systems in the early 20th century, waterborne outbreaks of communicable diseases including typhoid fever drastically decreased [35–37]. Improved sanitation measures such as having a basic pit latrine or a toilet connected to a septic tank curtail the direct contact between human waste and water or the environment. Yet in 2012, only 33% of the urban population in Uganda had access to adequate sanitation, an increase of only 1% since 1990, and 2% still practiced open defecation [38]. Although improving water and sanitation systems requires a substantial investment by the government, ultimately it is highly cost-effective in the reduction of many waterborne diseases [39]. Strengths and limitations A major limitation of our investigation was that, due to inadequate laboratory capacity to confirm a large number of cases early in the outbreak, and the need to rapidly identify the mode of transmission to inform effective interventions, we included non-laboratory confirmed cases in our case-control study. While a clinical case definition for typhoid fever cases can lead to misclassification, and is not recommended during nonoutbreak situations, during an outbreak such a case definition will often perform well, as measured by good positive and negative predictive values [40]. Also, the information on usual sources of water and food was based on self-reports, which could represent a source ofinformation bias. Another limitation is that data on mortality or on complications such as intestinal perforations were not collected. With over 10,000 cases it is likely that there were intestinal perforations and deaths but no surveillance for those outcomes was done. A study is currently being conducted to assess these severe impacts of this outbreak. In addition, only a few water and juice samples were tested, which could explain why S. Typhi was not isolated during the environmental investigation. Also, water and juice samples were tested using the reagents and procedures that were available in the laboratory for testing clinical specimens for Salmonella, and not more conventional methods for evaluating the potential presence of fecal contamination in these types of samples. Conclusion In conclusion, this investigation revealed a large outbreak of typhoid fever that affected thousands of people in Kampala, Uganda, which appeared to have been caused by consuming contaminated water and local drinks made from it. To prevent future waterborne outbreaks, we recommended that the Kampala Capital City Authority, the MoH, the National Water and Sewerage Corporation, and partners invest in improving access to potable water, and safe sanitation and hygiene facilities [41].

What were the infrastructure complaints?

232

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

What happened?

233

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

What was the event?

234

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

When did this happen?

235

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

When did this event start?

236

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

What is the date of this event?

237

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

How long was the event?

238

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

How long did the event last?

239

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

In which street did this happen?

240

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

In which city did this happen?

241

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

In which region did this happen?

242

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

In which country did this happen?

243

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

Where did this happen?

244

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

What caused the event?

245

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

What caused the event?

246

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

What caused the contamination of the tap water with river water?

247

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

What was the cause of the event?

248

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

What was the cause of the event?

249

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

What source started the event?

250

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

How was the event first detected?

251

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

How many people were ill?

252

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

How many people were hospitalized?

253

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

How many people were dead?

254

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

Which contaminants or viruses or bacteria were found?

255

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

Which were the symptoms?

256

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

What did the patients have?

257

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

What were the first steps?

258

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

What did they do to control the problem?

259

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

What did the local authorities advise?

260

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

What were the control measures?

261

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

What type of samples were examined?

262

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

What did they test for in the samples?

263

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

What is the concentration of the pathogens?

264

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

What steps were taken to restore the problem?

265

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

What was done to fix the problem?

266

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

What could have been done to prevent the event?

267

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

What could have been done to prevent the event?

268

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

What could have been done to prevent the event?

269

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

How to prevent this?

270

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

How to prevent this?

271

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

What were the investigation steps?

272

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

What were the investigation steps?

273

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

What did the investigation find?

274

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

How was the infrastructure affected?

275

A large community outbreak of gastroenteritis associated with consumption of drinking water contaminated by river water, Belgium, 2010

SUMMARY On 6 December 2010 a fire in Hemiksem, Belgium, was extinguished by the fire brigade with both river water and tap water. Local physicians were asked to report all cases of gastroenteritis. We conducted a retrospective cohort study among 1000 randomly selected households. We performed a statistical and geospatial analysis. Human stool samples, tap water and river water were tested for pathogens. Of the 1185 persons living in the 528 responding households, 222 (18·7%) reported symptoms of gastroenteritis during the time period 6–13 December. Drinking tap water was significantly associated with an increased risk for gastroenteritis (relative risk 3·67, 95% confidence interval 2·86–4·70) as was place of residence. Campylobacter sp. (2/56), norovirus GI and GII (11/56), rotavirus (1/56) and Giardia lamblia (3/56) were detected in stool samples. Tap water samples tested positive for faecal indicator bacteria and protozoa. The results support the hypothesis that a point-source contamination of the tap water with river water was the cause of the multi-pathogen waterborne outbreak. INTRODUCTION Tap water contaminated by gastrointestinal pathogens remains an important cause of gastrointestinal disease. In the European Union 86 enteric disease outbreaks associated with public drinking water supplies were reported from 1990 to 2004. In the USA 780 outbreaks were associated with drinking water from 1971 to 2006 [2]. Outbreaks are hard to detect and the number of cases associated with an outbreak varies. Waterborne disease is not limited to outbreaks. Sporadic cases probably represent a greater proportion of waterborne disease than cases related to outbreaks [1]. At the beginning of the 20th century, bacteria were most often identified as the cause of waterborne outbreaks. Nowadays viral and protozoal pathogens are commonly reported. Worldwide, 199 parasitic protozoa outbreaks with waterborne transmission were publicized during 2004–2010 [3]. Protozoa are also responsible for endemic disease associated with tap water [3–5]. A characteristic of pathogens associated with waterborne disease is a low infectious dose [6]. The bacteria most often found in North America and Europe in recent waterborne outbreaks are Campylobacter sp. and Escherichia coli [7–9]. In 2005 the most predominant parasitic protozoa isolated in waterborne outbreaks were Cryptosporidium, 60·3%, and Giardia lamblia, 35·2% [10]. Finally caliciviruses and viruses such as group A rotavirus have also been detected during waterborne outbreaks [11]. New methods for detecting norovirus (NoV) have resulted in increased detection of these pathogens [12–14]. Waterborne outbreaks have multiple causes, e.g. problems within the water system [15–17]. This can be failure of the disinfection system [11, 18], cracks in the service reservoir [19] or the mains [9], inappropriate connections between sewage- and drinking-water pipelines [20] or a pressure fall in the distribution system [8]. However, causes are not limited to problems with or failure of the system, e.g. the length of pipe run from the treatment works to the home, is correlated with the risk of disease [1]. Studies have described seasonal trends, with a higher proportion of waterborne outbreaks during spring and autumn, and associated these with (heavy) weather and agricultural activities [5, 21]. Often surface water is contaminated with runoff from regions with cattle and sheep [18]. On 6 December 2010 there was a fire in a textile factory in the centre of Hemiksem. Firefighters used water from two hydrants, connected to the tap water network, and from a unit ‘hydrosub’, which is used for pumping surface and river water. Car pumps collected water from both sources in a pressurized water tank. The hydrosub was connected to the river Vliet, a small river that flows to the river Scheldt. On 7 December three out of four routinely taken tap water samples in Hemiksem and Schelle suggested faecal contamination. Local general practitioners (GPs) reported an increase in consultations for gastroenteritis on 8 December. On 9 December the residents were advised not to consume or use tap water. The water avoidance notice was lifted on 20 December 2010. Tap water in the neighbouring municipalities of Hemiksem and Schelle (n = 18 620 residents), is supplied by one water supply company that uses purified ground water. The company takes care of collecting, cleaning and distributing drinking water. The surveillance, done according to the specific legislation, is a responsibility of the water company. Belgium has no specific surveillance system on waterborne outbreaks. We conducted an epidemiological study to describe the size and identify the source of the outbreak. METHODS The study population comprised of all the residents of Hemiksem and Schelle. We included no other municipalities given that elsewhere no faecal contamination in routinely taken samples of drinking water had been reported, and no increase of gastroenteritis had been registered. Considering that symptoms of gastroenteritis can be very mild, we combined a physicians' case-finding survey with a randomly sampled survey among the subscribers of the water company in Hemiksem and Schelle. Case reporting by physicians On 9 December all 18 local GPs were asked to report cases of gastroenteritis. Additionally, the emergency department of a neighbouring hospital was contacted and asked to report any resident of Hemiksem or Schelle who presented with gastroenteritis. Inclusion criteria were patients who lived in Hemiksem or Schelle, and who had symptoms of diarrhoea (⩾3 loose stools per 24 h) or vomiting from 6 to 13 December 2010. Retrospective cohort study A retrospective cohort study was conducted among a randomly selected sample of 1000 households from Hemiksem and Schelle. The sample was selected from a list of customers of the water supply company. A household was defined as all persons living at the same address. Through a postal survey every household member was asked for his personal usage of tap water (consumption, cooking, teeth brushing, cleaning), the usage of other water (ground water, bottled water), symptoms, with a focus on gastrointestinal symptoms, and the onset date of these symptoms and treatments. Any person reporting diarrhoea (⩾3 loose stools per 24 h) and/or vomiting between 6 and 13 December 2010, was defined as a case. Patients with an onset of diarrhoea or vomiting between 14 and 31 December 2010 were considered as late or secondary cases. Non-cases were people who did not report any symptoms. Households in which symptoms of gastroenteritis were reported prior to 6 December 2010 and persons who spent time abroad in the week prior to the outbreak were excluded from the study. We did not correct for a baseline number of gastroenteritis cases as no adequate incidence data for gastroenteritis was present for this region and time period. Based on the address of the respondent, the shortest distance, in metres, to the site of the fire was calculated. Descriptive statistics were calculated for the retrospective cohort study. We performed univariate analysis (cross tables, χ2 tests, linear trend analysis, relative risk) and multivariate analysis (Poisson regression). P < 0·05 was considered statistically significant. Variables associated with the outcome at P < 0·20 in univariate analysis were introduced in the multivariate regression model. The final multivariate model was built using backwards elimination [22]. The data were entered using EpiData v. 3.1 (EpiData Association, Denmark) and analysed using SAS v. 9.2 (SAS Institute Inc., USA) and R 2.14 (R Foundation, Austria). An additional analysis focusing on the geographical distribution of the cases was performed using the R package ‘sparr’ [23]. With this software the spatial density of the cases and the spatial density of non-cases was compared. Microbiological study Patients included in the physicians' survey were contacted and asked to provide a stool sample. These stool samples were tested for pathogenic gastrointestinal bacteria on culture media for isolation of enteropathogens (MacConkey agar, XLD and CIN agar) in a laboratory with accreditation to ISO 15 189. Antigen tests were used to test for Cryptosporidium and G. lamblia (Xpect Immunochromatographic Assays, Oxoid, UK). For the detection of NoV two methods were used. Three stool samples were send to the Scientific Institute of Public Health, Brussels, where they used a real-time RT–PCR for the detection and differentiation of the two most important human genogroups of NoV, GI and GII. This method is described in ISO/TS 15 216–1: 2013. The other 53 stool samples were send to the Clinical Virology Laboratory, Leuven and were analysed with a NoV RT–PCR. Two pairs of specific primers G1SKF/G1SKR and COG2F/G2SKR, that amplify the capsid gene, were used to respectively detect NoV GI (330 bp) and NoV GII (387 bp) [24]. An in-house RT–PCR was developed for detection of astrovirus. The primer set used to amplify 272 bp in the capsid protein gene was HASTV-F: ACAGAAGAGCAACTCCATCGC and HASTV-R: TGACACCYTGTTTCCTGAGTTG. A RT–PCR was used for rotavirus and adenovirus detection as described previously [25, 26]. The analyses of the tap water were performed by the water supply company. Samples were tested for bacterial indicators of faecal contamination: non-specific coliforms, Escherichia coli, Clostridium sp. and Pseudomonas aeruginosa. In two river water samples protozoa were isolated by the department of Parasitology of the University of Ghent. RESULTS Physicians' case-finding study All 18 GPs participated in the study. We included 603 patients; 326 (54%) were men and 277 (46%) women. The age ranged from 1 to 91 years with a median age of 36 years. Three hundred and ninety-seven (66%) cases lived in Hemiksem, and 206 (34%) in Schelle. Consultations for gastroenteritis peaked on 8 December 2010. Six patients were hospitalized. One patient, aged 91 years, died due to intestinal bleeding on 12 December after being admitted to the hospital for gastroenteritis on 9 December 2010. Retrospective cohort study The questionnaire was distributed on 20 December. The response rate was 52·8% (n = 528). The responses included information on a total of 1185 household members. This is 6·6% of the total population (n = 18 620) of Hemiksem and Schelle (December 2010). The respondents' age ranged from 0 to 99 years and the median age was 39 years. Gender was balanced with 48·1% males and 51·9% females. Out of all respondents, 37% regularly drank tap water, with an average of 3·7 glasses a day, 99% used tap water to brush their teeth, and 98% used tap water to wash vegetables and fruit. Gastrointestinal symptoms were reported by 36·5% (n = 432). The mean duration of illness was 4·1 days. Diarrhoea was the most frequently reported symptom, present in 65·4% of those reporting symptoms, fever in 15%, vomiting in 39% and nausea in 60%. Thirty-four respondents reported symptoms prior to 6 December. A total of 222 [18·7%, 95% confidence interval (CI) 16·4–20·9] persons met the case definition (onset of symptoms 6–13 December) and 176 were late or secondary cases (onset after 13 December). The incidence of gastroenteritis followed a steep incline from 6 to 8 December and peaked on 8 December (Fig. 1). The density of the cases (n = 222) over non-cases (n = 753) shows the highest proportion of cases slightly north of the site of the fire (Fig. 2). The density of self-reported ‘tap-water’ drinkers is compared to the density of ‘non-tap-water’ drinkers (see Fig. 3). A higher proportion of tap-water drinkers was observed east of Hemiksem. Univariate analysis Drinking tap water was associated with gastrointestinal symptoms [relative risk (RR) 2·28, 95% CI 1·94–2·67] (Table 1). The relative risk was higher for cases (RR 3·67, 95% CI 2·86–4·70) than for secondary or late cases (RR 1·91, 95% CI 1·47–2·49) (Table 1). The distance from the home address of the respondent to the site of the fire was significantly shorter for cases compared to non-cases. Other variables such as gender, household size, age and the presence of young children (<12 years) in the family were not significant (Table 2). Multivariate analysis The independent variables included in the multivariate model were: ‘glasses of tap water a day’, ‘location’ and ‘gender’. Location (distance from the site of the fire) as well as drinking tap water (glasses per day) were significantly (P < 0·05) associated with outcome in a Poisson regression model. For each glass of tap water consumed the risk was augmented by 21% (RR 1·21, 95% CI 1·16–1·26). For each 250 m that families lived further away from the site of the fire the risk diminished by 8·4% (RR 0·92, 95% CI 0·84–0·99) (Table 3). Microbiological patient data Sixteen of 56 stool samples were diagnosed with a pathogen. In the stool sample of 14 patients a single pathogen was detected: Campylobacter sp. was detected in one, NoV GI in two and NoV GII in seven patients. Rotavirus was detected in one patient and G. lamblia was detected in three samples. Two patients were diagnosed with a multi-pathogen infection: one patient with a double infection: NoV GI and GII, and one patient with a mixed infection with Campylobacter jejuni, NoV GI and GII. Environmental study Between 8 and 25 December 2010, 625 water samples were analysed in Hemiksem and Schelle. In Hemiksem, high densities of faecal pollution indicators were detected in tap water samples: >200 c.f.u./100 ml for E. coli and >1100 c.f.u./100 ml for enterococci. Parasites, G. lamblia and Cryptosporidium sp. were isolated from samples taken from the river. Control measures There were different authorities involved in controlling the outbreak and informing the population; the local municipal authorities, the environmental health team and the drinking water company. Bottled potable water was distributed to the population. Sanitation of water pipes (flushing, disinfection) was performed from 9 December 2010 to 25 December 2010. Our research pointed towards an incident during the firefighting on 6 December 2010 as the cause of the contamination. Additional research eliminated other possible reasons by checking installations in the surrounding area. An investigation into the actions taken by firefighters pointed towards the absence of reflux valves as the most likely cause. Once this hazard was identified, a warning was sent to other fire brigades in Belgium to avoid similar incidents. DISCUSSION We have described a large community outbreak of gastroenteritis associated with the consumption of water contaminated by river water during fire extinguishing. This is the first large waterborne outbreak described in Belgium. Waterborne disease, especially cryptosporidiosis, has been reported as a ‘work-related disease exposure for firefighters’, but it is the first time a community outbreak has been described in which the probable cause was associated with firefighting [27]. Our study estimated an attack rate of gastrointestinal infections of 18·7% without correction for baseline illness. The attack rate in our study can be explained by the high doses of pathogens in the tap water, a high daily consumption of tap water, a period of 3 days between the contamination and the water avoidance notice, and the low infectious dose of some pathogens. Attack rates in comparable studies vary and can be quite high, e.g. 51·4% in a tourist resort and new housing estate, 53% in a Finnish town [15, 20, 21, 28]. The number of cases per outbreak has been associated to the pathogens involved. Giardia- and Cryptosporidium-associated outbreaks have the lowest mean number of cases per outbreak (116 and 177, respectively). Viral outbreaks and outbreaks associated with Campylobacter sp. have the highest mean number of cases per outbreak (1545 and 1802, respectively) [1]. In a multi-pathogen outbreak, as we have described, one can expect an even higher number of cases. The number of hospitalizations was limited, but gastrointestinal disease can be associated with severe illness, especially in hig- risk patients [6] such as AIDS patients [29]. We found an unadjusted relative risk of 3·67 and the dose–response relationship was significant. A threefold increased risk for illness with the consumption of tap water is commonly associated with waterborne outbreaks [13, 16, 18, 19]. In E. coli and Campylobacter sp. outbreaks with relative risks of 11 were reported [7, 9]. In waterborne outbreaks, the strength of evidence implicating water as the cause of an outbreak is determined on the basis of findings from epidemiological and microbiological investigations. Tillett et al. developed a system of levels of evidence to link an outbreak to water [30]. We complement this by adding an important spatial component. We observed a steep increase in reported gastroenteritis among the survey's responders on the 6 December which indicates a sudden contamination. Microbiological investigations during this increase identified multiple pathogens in the stool samples of patients. This is consistent with the hypothesis of a waterborne outbreak rather than a community-wide person-to-person transmission of e.g. a NoV [31]. Furthermore, tap water was a likely mode of transmission as there was a significant association between gastroenteritis and the consumption of tap water. Enterobacteriaceae were found in the tap water. We linked the firefighting to the outbreak by spatial and temporal analysis. We documented the association between place of residence and the risk of gastroenteritis by comparing the ratio of cases (n = 222) over non-cases (n = 753) with the ratio of tap-water drinkers over non-tap-water drinkers. We observed that one peak in the ratio of cases over controls was not accompanied by a high ratio of tap-water drinkers over non-tap-water drinkers. This cluster was located slightly north of the site of the fire. We have no data on quantitative difference of pathogen load in the pipelines. A major weakness of this study is the incomplete microbiological investigation. Drinking water contaminated by sewage is known to result in mixed bacterial and viral infections and severe acute gastroenteritis regardless of the aetiological agents [17, 32]. Investigation of the tap water was limited to indicator bacteria for faecal contamination and a one-time detection of Cryptosporidium sp. and G. lamblia. Patients were not tested for protozoa. Detection of e.g. levels of serum antibody to G. lamblia or testing of patients for Cryptosporidium should be performed in this kind of outbreak investigation [33, 34]. No attempt was made to match the clinical and environmental pathogens. Pathogens from the environmental samples were not stored and therefore not available for further testing. Since the incubation period of several gastrointestinal pathogens can vary from hours to weeks, we took the entire month of December 2010 as study period. However, the difference between cases and late or secondary cases is artificial and only used as a means to better analyse the data. Long incubation periods and secondary infections by person-to-person transmission are hard to differentiate. Self-reported information on symptoms and water consumption is known to be biased, especially after media attention [35]. However, the media did not report on possible sources and the connection to the fire. The media only started to report on the outbreak after the water avoidance notice. We tried to collect objective information by contacting healthcare personnel. The questionnaire was sent rather late, on 20 December 2010, which could generate recall bias. Moreover, persons affected by the outbreak might be more inclined to respond to the survey leading to selection bias. No information was collected on the costs of this outbreak. Previous research shows that the costs for preventive measures clearly are smaller than the costs of a waterborne outbreak [36]. We did not investigate compliance with the water avoidance notice. A study in The Netherlands estimated compliance with a boil water advice at 81·8% [37]. Several improvements can be recommended such as the development of a surveillance system and a wider and more thorough microbiological investigation. Syndromic surveillance combined with water incident and consumer complaints data can be used for the timely detection of outbreaks, but this needs to be further evaluated [38]. More intense microbiological investigations are necessary both during standard controls and outbreaks. Previous studies have indicated that water-treatment technologies have become inadequate, and that a negative coliform test result does not guarantee that water is free from all pathogens, especially from protozoan agents [6]. Both enteric viruses, such as caliciviruses, and some protozoan agents, such as Cryptosporidium, are candidates for endemic transmission and outbreaks with these pathogens will go undetected with standard controls [39]. Furthermore, a more thorough microbiological investigation can be used to predict the likelihood of various transmission routes or vehicles [40]. This outbreak also highlights the need to rapidly connect an outbreak to its cause to reduce attack rates by implementing the correct measurements. This outbreak could have been detected quicker if additional surveillance systems had been in place. Rapid detection and intervention necessitates the collaboration between physicians, public health services, microbiologists and water providers.

What were the infrastructure complaints?

276

A large multi-pathogen waterborne community outbreak linked to faecal contamination of a groundwater system, France, 2000

A large waterborne outbreak of infection that occurred during August 2000 in a local community in France was investigated initially via a rapid survey of visits to local physicians. A retrospective cohort study was then conducted on a random cluster sample of residents. Of 709 residents interviewed, 202 (28.5%) were definite cases (at least three liquid stools ℠day or vomiting) and 62 (8.7%) were probable cases (less than three liquid stools ℠day or abdominal pain). Those who had drunk tap water had a threefold increased risk for illness (95% CI 2.4–4.0). The risk increased with the amount of water consumed (chi-square trend: p < 0.0001). Bacteriological analyses of stools were performed for 35 patients and virological analyses for 24 patients. Campylobacter coli, group A rotavirus and norovirus were detected in 31.5%, 71.0% and 21% of samples, respectively. An extensive environmental investigation concluded that a groundwater source to this community had probably been contaminated by agricultural run-off, and a failure in the chlorination system was identified. This is the first documented waterborne outbreak of infection involving human C. coli infections. A better understanding of the factors influencing campylobacter transmission between hosts is required. Keywords Campylobacter coli, epidemiology, gastroenteritis, norovirus, rotavirus, waterborne outbreak INTRODUCTION Contaminated drinking water causes extensive outbreaks of illness because of the large number of people served by water supply facilities. Detection of such outbreaks requires the identification of an increase in illness (usually the rate of gastrointestinal disease) in the exposed population and confirmation that water was the route of transmission [1]. Many countries have routine reporting systems for detecting foodborne and waterborne outbreaks of infection. However, in routine practice it is not easy to detect an increase in clinical cases and to link this increase to waterborne transmission; thus waterborne outbreaks are often unrecognised and underestimated [2]. Microbiological examination of water is more complex than examination of stools, and there is often a failure to detect pathogens in water. Norovirus, Cryptosporidium parvum, Giardia intestinalis and Campylobacter jejuni are the pathogens identified most frequently [3]. The present study describes a large waterborne community outbreak of infection with multiple pathogens, including Campylobacter coli. METHODS AND MATERIALS Background On 23 August 2000, a general practitioner (GP) informed the local district health department that 16 cases of acute gastroenteritis (AGE) had occurred among residents of a holiday camp in the Gourdon community since 21 August. Gourdon, which is located in the south-west of France, has a population of 4888 inhabitants, which increases by 50% in the July– August tourist season. Investigations at the holiday camp concluded that foodborne transmission was very unlikely, but found that tap water was highly contaminated with faecal coliforms. Preliminary information collected from physicians in Gourdon indicated an increase of the number of consultations for AGE. Gourdon tap water is supplied by two underground water facilities: facility A serving 3800 water consumers in Gourdon; and facility B serving 1200 water consumers in Gourdon and the surrounding communities. Water treatment in facility A consists of chlorination, which involves two automatic pumps working in rotation. The water then passes through two semi in-ground reservoir tanks (1100 m3 and 200 m3, respectively) before distribution. Facility B water is mixed with a fraction of facility A water in another 500 m3 semi in-ground reservoir tank before being distributed. This mixing with the treated water from facility A is intended to disinfect the water from facility B. Since the increase in AGE was noted only in Gourdon, contaminated tap water from facility A was suspected to be the origin of the outbreak. Therefore, the population of Gourdon was informed on 25 August to avoid the consumption of tap water or to boil it for 5 min. Survey of gastroenteritis-related medical consultations To confirm the outbreak, the 11 GPs in Gourdon and the local hospital emergency department were requested to provide the number of total daily visits and the number of cases of gastroenteritis cases occurring between 1 August and 11 September 2000. Retrospective cohort study A retrospective cohort study was performed on a sample of Gourdon households between 5 and 15 September 2000. Domestic households were selected randomly from the telephone directory. A definite case was defined as a Gourdon resident who had at least three liquid stools ℠day or vomiting between 1 August and 3 September, and who had been present in Gourdon for ‡ 1 day between 1 and 31 August; if less than three liquid stools ℠day or abdominal pain had occurred, the case was classified as probable. Exposure was defined as the consumption of tap water before 25 August 2000. A standardised telephone questionnaire was completed to obtain demographic data (gender, date of birth), food consumption (commercially prepared dishes, sea-food, raw milk, pastry), tap water consumption (before and after 25 August, daily number of glasses of water, consumption of boiled tap water), symptoms of gastroenteritis (date of onset, number of liquid stools ℠day, vomiting, nausea, abdominal pain, fever), medical care (in- or out-patient) and consequences on daily activities (sick leave, bed confinement) for each member of the selected household. Under the hypothesis that 60% of the population consumed tap water, an estimated 550 people were required for inclusion in the study to identify a relative risk (RR) of ‡ 3, with a 5% a risk and a power of 80%. Based on an average of two individuals per household (INSERM 1999 population census), 300 households were required for inclusion in the survey. The association between AGE and the consumption of tap water was assessed by calculating the risk ratio (RR) and 95% CI, taking into account a household cluster design effect. A dose-effect relationship was evaluated by analysing the RR trend for different categories of daily number of glasses of tap water consumed (chi-square trend). Proportions were compared using the corrected chi-square test, and means were compared using the Student test or Mann and Witney test. Data were analysed with Epi-Info 6.04c (CDC, Atlanta, GA, USA). The gender and age distribution of the cohort population were compared with that of the general population of Gourdon (INSERM 1999 population census) to assess whether it was representative. Under the hypothesis of representation, the total number of cases of gastroenteritis was estimated by applying the attack rate to the total population of Gourdon, including the tourist population (data provided by the Gourdon tourism office). The most likely period for contamination of the water distribution system was based on the epidemic curve and the minimal incubation period of the pathogens involved. Microbiological analyses of stools Stool samples were examined for the presence of Salmonella spp., Shigella spp., Staphylococcus aureus, Campylobacter spp., Yersinia enterocolitica, Escherichia coli, enterococci, rotavirus groups A and C, astrovirus, calicivirus, adenovirus types 40 and 41, enterovirus, hepatitis A virus and Cryptosporidum. Viruses were detected using an enzyme immunoassay for group A rotavirus, astrovirus and adenovirus 40 ℠41, and by RT-PCR for calicivirus, group C rotavirus, enterovirus and hepatitis A virus [4–6]. Campylobacter isolates were sent to the National Reference Centre for Campylobacter for identification and fingerprinting. Species identification was by means of phenotypic tests (API Campy; bioMe´rieux, Lyon, France), with use of PCR to differentiate Campylobacter jejuni from Camplylobacter coli [7]. Random amplified polymorphic DNA fingerprinting was performed with primer 3881, 5¢-AACGCGCAAC. Environmental investigation An environmental and microbiological investigation of the water distribution system was performed. The inspection concerned the water catchments, the water distribution and sewage system, agricultural practices and private households around the sources. Bacteriological (Escherichia coli, enterococci, faecal streptococci, thermotolerant coliforms, sulphite-reducing clostridia) and chemical (chlorine) analyses were performed on tap water and on the water sampled from the two groundwater sources. After 23 August, the water chlorine concentration and the presence of coliform bacteria were monitored daily at several sites of the water system before and after the resort treatment. Examinations for viruses (group A rotavirus, astrovirus, norovirus, enterovirus, hepatitis A virus) and Campylobacter spp. were undertaken on two 10-mL tap water samples collected inhouse on 23 August, and on two 10-mL samples from groundwater sources A and B on 25 August. Viruses were detected by RT-PCR, followed by hybridisation with a specific primer to improve sensitivity, while C. jejuni and C. coli were detected by PCR with specific primers. Group A rotavirus isolates detected in stools and in water were sequenced using Beg 9 and End 9 primers [8] and compared using the Infobiogene GCG software (http://www.infobiogen.fr). Bacteriological (revived 36C, revived 22C, thermotolerant coliforms, streptococci and faecal enterococci), physical (temperature, conductivity) and chemical (oxygen consumption, ammonium, nitrates, phosphorus) parameters from source A and the river were analysed. Trends were compared for an 8-week period between 21 December 2000 and 7 February 2001. RESULTS Epidemiology Between 1 August and 11 September 2000, 7104 Gourdon residents consulted a GP or the local hospital emergency department. Between 24 and 28 August, gastroenteritis accounted for 44% (479 ℠1097) of all medical visits, compared with 6% (126 ℠2185) during the first 15 days of August (Fig. 1). Among the 498 households contacted for the household survey, 198 were excluded (103 did not respond, 49 did not fulfil the inclusion criteria, 43 refused to participate, and three for unknown reasons). The questionnaire was completed for 300 households (709 individuals). Data were missing concerning gender for one individual and age for 11 individuals. Overall, 331 (46.7%; 95% CI 44.5–49.0) of respondents were male and 377 (53.2%; 95% CI 50.9–55.5) were female. The median age was 44 years (range < 1–94 years). The age and gender distribution of the sample was similar to that of the general population. During the outbreak period, 264 of the 709 respondents (attack rate (AR) 37.2%) had been ill, with 202 definite cases (AR 28.5%; 95% CI 24.6– 32.4) and 62 probable cases (AR 8.7%; 95% CI 6.5– 11.0). The AR was higher in females (31.8%) than in males (24.8%), and was highest in children aged < 6 years (42.1%, Table 1). The epidemic curve (Fig. 2) suggests that the outbreak started between 14 and 19 August. The shape of the epidemic curve in the retrospective cohort study was very similar to that based on the number of visits to GPs for gastroenteritis. Patients complained mainly of diarrhoea, abdominal pain and nausea (Table 2). No cases of bloody diarrhoea were notified. The mean duration of symptoms was 4.5 days (median 3, range 1–30 days). Among definite cases, 52% had visited a GP, 42% had taken sick leave for ‡ 1 day, 32% had required bed rest, and 2.9% (six patients) had been hospitalised. In total, 336 (47.4%) respondents had drunk tap water before 25 August, whereas 646 (91.1%) did not drink tap water after 25 August. The risk for illness, taking into account the sample cluster design effect of 1.4, was three-fold greater (95% CI 2.2–4.1) for those who had drunk tap water before 25 August, and increased with the amount of water consumed (chi-square trend: p < 0.0001; Table 3). The risk of gastroenteritis associated with drinking tap water did not change substantially after adjustment for age (RRMH 3.3, 95% CI 2.5–4.4), and there was no modification effect according to age. Food consumption was not associated with illness. Applying the AR of 37% to the total population of Gourdon exposed to the water supply (4888 residents + 2200 tourists), the number of individuals affected by this outbreak was estimated to be 2600 (95% CI 2400–2900), assuming an exposure to tap water and a risk for gastroenteritis among the tourists similar to that among the inhabitants. Microbiological stool analyses Among the 35 stool samples analysed for bacteria, 11 (31.5%) were positive for C. coli. Twenty-four of the 35 samples were also tested for viruses: rotavirus A was detected in 17 (71.0%) and norovirus in five (21.0%) samples, while nine (37.5%) samples were co-infected (four C. coli + rotavirus; one C. coli + norovirus; three rotavirus + norovirus; and one C. coli + rotavirus + norovirus). Of the 11 C. coli isolates, six were characterised as biotype 2, and were identical following RAPD analysis, while the remaining five were of biotype 1. Rotavirus typing by RTPCR identified a single genotype P-(8), G1. Molecular characterisation of the norovirus isolates identified two different patterns, with three isolates being identical and close to the Saragota strain (genogroup I) and two close to the new GGIIb variant (genogroup II). Environmental investigation All of the water samples from source B were negative, regardless of the site of the distribution system tested. Analyses of the water originating from source A, sampled after leaving the water-works on 23 and 25 August, revealed high concentrations of E. coli (2263 / 100 mL), enterococci (415 / 100 mL), total coliforms (90 / 100 mL), faecal streptococci (136 / 100 mL) and sulphite-reducing clostridia (10 / 100 mL), and an absence of chlorination. Rotavirus A genotype G1 was detected in the crude water sampled from source A. The analyses for enterovirus, astrovirus, norovirus, hepatitis A virus and Campylobacter were negative. Source A was situated in a valley containing a river, below a hamlet of 15 dwellings, including farms breeding cattle, pork and sheep. The source was not supposed to be supplied by the river. The river resulted from the convergence of about ten different streams, with one originating close to a slaughterhouse, and was exposed to agricultural run-off and manure effluent; sheep and cattle grazed in meadows bordering the river, and a large amount of sewage from water treatment facilities was sprayed on to land close to the river. Cowsheds were close to the water catchment area. The immediate protected area around source A was too small and littered with building rubbish. Air vents without protection in the door made the source accessible to animals and insects. Bacteriological analyses of the water samples from source A and the river revealed that the contamination was of both human and animal origin, with a higher concentration in the river. The levels and trends of the chemical parameters of source A were similar to those of the river, strongly suggesting that the river was connected to source A (Fig. 3). A retrospective review of routine data collected between 1991 and 2000 showed irregularities in the maintenance of the water distribution system and the water treatment plant, particularly during the month of August. In the hamlet, the sewage system of the 15 households did not comply with sanitary rules. All but three ejected waste either into a cesspool or superficially without treatment. Three regions of gene 9 of the rotavirus detected in stools and water were compared. Sequencing showed that the strains were closely related but not identical. Assuming a likely onset of the outbreak between 14 and 19 August, and taking into account a minimum incubation period of 2 days for rotavirus A, norovirus and C. coli, the contamination of the water distribution system probably occurred between 12 and 17 August. Contamination probably persisted until 28 August, as intensive chlorination, beginning on 25 August, was effective only from 28 August onwards (Fig. 2). DISCUSSION The present descriptive and retrospective cohort studies confirmed that the outbreak of infection in the community of Gourdon was related to the consumption of tap water and affected 2600 individuals. The outbreak involved multiple pathogens, with group A rotavirus and C. coli being the most frequent, followed by norovirus. Results of environmental investigations suggested that water contamination was related to the contamination of the source, combined with a failure of the water chlorination system. The association with tap water consumption, the dose-effect relationship, and the fact that multiple pathogens were involved, were all strongly in favour of tap water as the origin of infection. Moreover, the number of cases decreased dramatically after the water restriction advice and the implementation of control measures on the water supply. No food products explored in the study were associated with illness. An AR of 14.5% and 31% was observed among adults and children aged < 6 years, respectively, who reported that they had not consumed tap water. This relatively high rate could be explained by several phenomena. Many individuals may have consumed tap water without being aware of it, e.g., in drinks or ice cubes consumed outside of the home, via raw fruits and vegetables, dishes prepared with unboiled water, or while cleaning their teeth. It can be hypothesised that people who became ill without being aware of having consumed tap water were exposed to small amounts of water and had been subjected to a small infective dose. In addition, children, more than adults, may have been contaminated through person-to-person transmission (e.g., family contact), which is the most common route of transmission for rotavirus and calicivirus. Waterborne contamination can affect large numbers of individuals [3,9–11]. In the present study, assuming a 50% increase in the population during the tourist season, the number of affected people was estimated to be between 2400 and 2900. The cohort study showed that almost half (46.5%) of the cases consulted a GP. Applying this proportion to the estimated number of cases yielded an estimated 1116–1349 ill individuals who consulted a GP for gastroenteritis, similar to the number of consultations for gastroenteritis actually reported by the GPs (n = 1037). The stool sample analyses revealed the presence of multiple pathogens (rotavirus A, norovirus genogroups I and II and C. coli) as single or co-infections. Except for C. coli, these pathogens have been implicated in numerous waterborne outbreaks, suggesting an influx of sewage into the water supply system [10,12–18]. Rotaviruses are known to be responsible for large epidemics during winter seasons, affecting particularly children aged < 3 years. These epidemics are caused by the faecal route of contamination (personto-person transmission). The high proportion of stools positive for rotavirus (71%) is uncommon in waterborne outbreaks and probably reflects massive contamination [15,16,19]. Different molecular sequences of rotavirus A genotype G1 found in stools and water indicated a human faecal source of infection. In the present study, contamination of water by sewage could be linked with the large amount of sewage from the water purification plant sprayed on to land close to the river and the water catchments. In this outbreak, C. coli, which was identified in 32% of stools, was the second most frequent pathogen. To our knowledge, human C. coli contamination has not been described previously as a causative agent in waterborne outbreaks, unlike C. jejuni [11,20–25], although C. coli has been isolated from water [26]. The main symptoms observed (diarrhoea, abdominal pain, nausea) and the mean duration of the disease (4.5 days) were compatible with a Campylobacter infection. No bloody diarrhoea was reported, although bloody diarrhoea is not constant in Campylobacter infections [27]. The high proportion of individuals who had to take sick leave, and the fact that six individuals were hospitalised, suggested a relative severity of symptoms, which is much more compatible with Campylobacter infection than with virus infections. Among the hospitalised patients, 60% were infected with C. coli. Among patients whose stools were analysed and whose age was known, cases infected only with rotavirus were younger (mean age 49 years; range, 5 months to 95 years) than cases infected only with Campylobacter (mean age 66 years; range 23–88 years) or co-infected with rotavirus plus Campylobacter (mean age 69 years; range 2–91 years). Moreover, the proportion (52%) of medical visits was unusually high compared with the proportion of visits observed during other waterborne outbreaks [13,16,28]. Campylobacter needs specific atmospheric conditions for growth and survival and its detection can be very difficult [29,30]. Because of a rapid physiological transformation to a viable, though non-cultivable, state, it has been suggested that conventional culture methods often fail [31]. Nevertheless, in the present study, the failure to detect C. coli in water, and the detection of noroviruses in stools, could be caused by the small (10 mL) water samples examined. Ha¨nninen et al. [30] proposed that the volume used for detection of suspected pathogens in drinking water should be at least 8–10 L. Groundwater is considered widely to be microbiologically clean, and is usually used for drinking without treatment. However, groundwater can be contaminated by the environment [26,32]. Stanley et al. [33] showed that Campylobacter can occur in groundwater, with hydrological evidence suggesting that the source of contamination in their study was a dairy farm situated within the hydrological catchment area of the groundwater, which was confirmed when identical strains of C. jejuni were isolated from groundwater and the dairy herd. In the present study, Campylobacter was not identified from groundwater, and dairy herds and other domestic animals in farms around the source were not investigated. However, results of bacterial, chemical and physical water analyses suggested strongly that a river contaminated by agricultural run-off from surrounding sheep and cattle grazing meadows was contaminating the source. In addition, the immediate area of the source was insufficiently protected and vulnerable to the environment. Wild birds, rodents, pigs and cows are all hosts for C. coli and could contaminate surface water easily [26,27,34,35]. Campylobacters are sensitive to many environmental factors and to chlorine [33,36,37], but a failure of the chlorination system could have resulted in the presence of a high concentration of C. coli. Waterborne outbreaks are often caused by several factors that act together [18,20,23,38]. In some outbreaks, heavy rainfall has been responsible for increasing agricultural run-off into rivers, although this did not occur before the outbreak in Gourdon. This underlies the need for increased frequency of routine water quality monitoring, chlorination of the water supply according to environmental conditions, and implementation of preventive measures. This outbreak raised the question of the delay in raising the alert. An outbreak can be recognised by: (i) the poor quality of water detected during regular controls; (ii) complaints about water quality; (iii) an incident in the supply and ℠or sewage water system during maintenance work; and (iv) an increased number of cases of gastroenteritis. In most instances, waterborne outbreaks are suspected when there are already a large number of declared cases [3,12,18,28,38,39]. In the present outbreak, the initial warning was given by a GP who notified officials 5–10 days after the assumed date of onset. This highlights the role of GPs in the detection of an unusual event, particularly during the summer, when an increase in acute gastroenteritis is less expected than during winter [40,41]. In addition, a routine water quality monitoring system able to detect contamination of drinking water before the occurrence of illness and complaints about the water quality should be used as an early warning source. The retrospective cohort study of Gourdon residents allowed the scale and severity of the outbreak to be quantified precisely and the hypothesis that tap water was the vehicle of the outbreak to be tested. When provided with a cohort sample representative of the study population, the number of people affected can be estimated. However, as the exposure of the population was high, a relatively large sample size was needed to test the association between the disease and the exposure. Because of misclassification of exposure (unrecognised consumption of tap water), the attributable risk is probably underestimated. In the absence of obvious alternative explanations, water contamination is the most likely common cause of an explosive outbreak of AGE occurring in a large population. Tillett et al. [42] proposed categorising levels of evidence for waterborne outbreaks. A descriptive study suggesting a water-related outbreak, together with the identification of the same pathogen in stools and water, should be sufficient to conclude that an outbreak is associated strongly with water consumption. Faecal indicators of water quality are currently insufficiently sensitive for some pathogens, notably viruses and parasites (e.g., Cryptosporidium). Therefore, in the absence of evidence from an analytical study to demonstrate an association, the link between illness and water can only be classified as probable or possible. However, total coliform counts are good indicators of faecal contamination with other pathogens (e.g., rotaviruses, E. coli, Campylobacter) [43]. In conclusion, the waterborne outbreak described in the present study was most probably caused by faecal contamination of source A by groundwater, resulting from deficiencies in the maintenance of the water distribution system. Corrective measures were implemented based on the results of the investigation. To our knowledge, this is the first documented waterborne outbreak involving human C. coli infection. In-depth environmental investigations to identify the origin of water contamination are important in order to better understand microbial transmission from the environment and to implement appropriate controls and preventive measures. In addition, surveys on vehicles and vectors that transmit Campylobacter between hosts are crucial in order to better understand the epidemiology of these organisms.

What happened?

277

A large multi-pathogen waterborne community outbreak linked to faecal contamination of a groundwater system, France, 2000

A large waterborne outbreak of infection that occurred during August 2000 in a local community in France was investigated initially via a rapid survey of visits to local physicians. A retrospective cohort study was then conducted on a random cluster sample of residents. Of 709 residents interviewed, 202 (28.5%) were definite cases (at least three liquid stools ℠day or vomiting) and 62 (8.7%) were probable cases (less than three liquid stools ℠day or abdominal pain). Those who had drunk tap water had a threefold increased risk for illness (95% CI 2.4–4.0). The risk increased with the amount of water consumed (chi-square trend: p < 0.0001). Bacteriological analyses of stools were performed for 35 patients and virological analyses for 24 patients. Campylobacter coli, group A rotavirus and norovirus were detected in 31.5%, 71.0% and 21% of samples, respectively. An extensive environmental investigation concluded that a groundwater source to this community had probably been contaminated by agricultural run-off, and a failure in the chlorination system was identified. This is the first documented waterborne outbreak of infection involving human C. coli infections. A better understanding of the factors influencing campylobacter transmission between hosts is required. Keywords Campylobacter coli, epidemiology, gastroenteritis, norovirus, rotavirus, waterborne outbreak INTRODUCTION Contaminated drinking water causes extensive outbreaks of illness because of the large number of people served by water supply facilities. Detection of such outbreaks requires the identification of an increase in illness (usually the rate of gastrointestinal disease) in the exposed population and confirmation that water was the route of transmission [1]. Many countries have routine reporting systems for detecting foodborne and waterborne outbreaks of infection. However, in routine practice it is not easy to detect an increase in clinical cases and to link this increase to waterborne transmission; thus waterborne outbreaks are often unrecognised and underestimated [2]. Microbiological examination of water is more complex than examination of stools, and there is often a failure to detect pathogens in water. Norovirus, Cryptosporidium parvum, Giardia intestinalis and Campylobacter jejuni are the pathogens identified most frequently [3]. The present study describes a large waterborne community outbreak of infection with multiple pathogens, including Campylobacter coli. METHODS AND MATERIALS Background On 23 August 2000, a general practitioner (GP) informed the local district health department that 16 cases of acute gastroenteritis (AGE) had occurred among residents of a holiday camp in the Gourdon community since 21 August. Gourdon, which is located in the south-west of France, has a population of 4888 inhabitants, which increases by 50% in the July– August tourist season. Investigations at the holiday camp concluded that foodborne transmission was very unlikely, but found that tap water was highly contaminated with faecal coliforms. Preliminary information collected from physicians in Gourdon indicated an increase of the number of consultations for AGE. Gourdon tap water is supplied by two underground water facilities: facility A serving 3800 water consumers in Gourdon; and facility B serving 1200 water consumers in Gourdon and the surrounding communities. Water treatment in facility A consists of chlorination, which involves two automatic pumps working in rotation. The water then passes through two semi in-ground reservoir tanks (1100 m3 and 200 m3, respectively) before distribution. Facility B water is mixed with a fraction of facility A water in another 500 m3 semi in-ground reservoir tank before being distributed. This mixing with the treated water from facility A is intended to disinfect the water from facility B. Since the increase in AGE was noted only in Gourdon, contaminated tap water from facility A was suspected to be the origin of the outbreak. Therefore, the population of Gourdon was informed on 25 August to avoid the consumption of tap water or to boil it for 5 min. Survey of gastroenteritis-related medical consultations To confirm the outbreak, the 11 GPs in Gourdon and the local hospital emergency department were requested to provide the number of total daily visits and the number of cases of gastroenteritis cases occurring between 1 August and 11 September 2000. Retrospective cohort study A retrospective cohort study was performed on a sample of Gourdon households between 5 and 15 September 2000. Domestic households were selected randomly from the telephone directory. A definite case was defined as a Gourdon resident who had at least three liquid stools ℠day or vomiting between 1 August and 3 September, and who had been present in Gourdon for ‡ 1 day between 1 and 31 August; if less than three liquid stools ℠day or abdominal pain had occurred, the case was classified as probable. Exposure was defined as the consumption of tap water before 25 August 2000. A standardised telephone questionnaire was completed to obtain demographic data (gender, date of birth), food consumption (commercially prepared dishes, sea-food, raw milk, pastry), tap water consumption (before and after 25 August, daily number of glasses of water, consumption of boiled tap water), symptoms of gastroenteritis (date of onset, number of liquid stools ℠day, vomiting, nausea, abdominal pain, fever), medical care (in- or out-patient) and consequences on daily activities (sick leave, bed confinement) for each member of the selected household. Under the hypothesis that 60% of the population consumed tap water, an estimated 550 people were required for inclusion in the study to identify a relative risk (RR) of ‡ 3, with a 5% a risk and a power of 80%. Based on an average of two individuals per household (INSERM 1999 population census), 300 households were required for inclusion in the survey. The association between AGE and the consumption of tap water was assessed by calculating the risk ratio (RR) and 95% CI, taking into account a household cluster design effect. A dose-effect relationship was evaluated by analysing the RR trend for different categories of daily number of glasses of tap water consumed (chi-square trend). Proportions were compared using the corrected chi-square test, and means were compared using the Student test or Mann and Witney test. Data were analysed with Epi-Info 6.04c (CDC, Atlanta, GA, USA). The gender and age distribution of the cohort population were compared with that of the general population of Gourdon (INSERM 1999 population census) to assess whether it was representative. Under the hypothesis of representation, the total number of cases of gastroenteritis was estimated by applying the attack rate to the total population of Gourdon, including the tourist population (data provided by the Gourdon tourism office). The most likely period for contamination of the water distribution system was based on the epidemic curve and the minimal incubation period of the pathogens involved. Microbiological analyses of stools Stool samples were examined for the presence of Salmonella spp., Shigella spp., Staphylococcus aureus, Campylobacter spp., Yersinia enterocolitica, Escherichia coli, enterococci, rotavirus groups A and C, astrovirus, calicivirus, adenovirus types 40 and 41, enterovirus, hepatitis A virus and Cryptosporidum. Viruses were detected using an enzyme immunoassay for group A rotavirus, astrovirus and adenovirus 40 ℠41, and by RT-PCR for calicivirus, group C rotavirus, enterovirus and hepatitis A virus [4–6]. Campylobacter isolates were sent to the National Reference Centre for Campylobacter for identification and fingerprinting. Species identification was by means of phenotypic tests (API Campy; bioMe´rieux, Lyon, France), with use of PCR to differentiate Campylobacter jejuni from Camplylobacter coli [7]. Random amplified polymorphic DNA fingerprinting was performed with primer 3881, 5¢-AACGCGCAAC. Environmental investigation An environmental and microbiological investigation of the water distribution system was performed. The inspection concerned the water catchments, the water distribution and sewage system, agricultural practices and private households around the sources. Bacteriological (Escherichia coli, enterococci, faecal streptococci, thermotolerant coliforms, sulphite-reducing clostridia) and chemical (chlorine) analyses were performed on tap water and on the water sampled from the two groundwater sources. After 23 August, the water chlorine concentration and the presence of coliform bacteria were monitored daily at several sites of the water system before and after the resort treatment. Examinations for viruses (group A rotavirus, astrovirus, norovirus, enterovirus, hepatitis A virus) and Campylobacter spp. were undertaken on two 10-mL tap water samples collected inhouse on 23 August, and on two 10-mL samples from groundwater sources A and B on 25 August. Viruses were detected by RT-PCR, followed by hybridisation with a specific primer to improve sensitivity, while C. jejuni and C. coli were detected by PCR with specific primers. Group A rotavirus isolates detected in stools and in water were sequenced using Beg 9 and End 9 primers [8] and compared using the Infobiogene GCG software (http://www.infobiogen.fr). Bacteriological (revived 36C, revived 22C, thermotolerant coliforms, streptococci and faecal enterococci), physical (temperature, conductivity) and chemical (oxygen consumption, ammonium, nitrates, phosphorus) parameters from source A and the river were analysed. Trends were compared for an 8-week period between 21 December 2000 and 7 February 2001. RESULTS Epidemiology Between 1 August and 11 September 2000, 7104 Gourdon residents consulted a GP or the local hospital emergency department. Between 24 and 28 August, gastroenteritis accounted for 44% (479 ℠1097) of all medical visits, compared with 6% (126 ℠2185) during the first 15 days of August (Fig. 1). Among the 498 households contacted for the household survey, 198 were excluded (103 did not respond, 49 did not fulfil the inclusion criteria, 43 refused to participate, and three for unknown reasons). The questionnaire was completed for 300 households (709 individuals). Data were missing concerning gender for one individual and age for 11 individuals. Overall, 331 (46.7%; 95% CI 44.5–49.0) of respondents were male and 377 (53.2%; 95% CI 50.9–55.5) were female. The median age was 44 years (range < 1–94 years). The age and gender distribution of the sample was similar to that of the general population. During the outbreak period, 264 of the 709 respondents (attack rate (AR) 37.2%) had been ill, with 202 definite cases (AR 28.5%; 95% CI 24.6– 32.4) and 62 probable cases (AR 8.7%; 95% CI 6.5– 11.0). The AR was higher in females (31.8%) than in males (24.8%), and was highest in children aged < 6 years (42.1%, Table 1). The epidemic curve (Fig. 2) suggests that the outbreak started between 14 and 19 August. The shape of the epidemic curve in the retrospective cohort study was very similar to that based on the number of visits to GPs for gastroenteritis. Patients complained mainly of diarrhoea, abdominal pain and nausea (Table 2). No cases of bloody diarrhoea were notified. The mean duration of symptoms was 4.5 days (median 3, range 1–30 days). Among definite cases, 52% had visited a GP, 42% had taken sick leave for ‡ 1 day, 32% had required bed rest, and 2.9% (six patients) had been hospitalised. In total, 336 (47.4%) respondents had drunk tap water before 25 August, whereas 646 (91.1%) did not drink tap water after 25 August. The risk for illness, taking into account the sample cluster design effect of 1.4, was three-fold greater (95% CI 2.2–4.1) for those who had drunk tap water before 25 August, and increased with the amount of water consumed (chi-square trend: p < 0.0001; Table 3). The risk of gastroenteritis associated with drinking tap water did not change substantially after adjustment for age (RRMH 3.3, 95% CI 2.5–4.4), and there was no modification effect according to age. Food consumption was not associated with illness. Applying the AR of 37% to the total population of Gourdon exposed to the water supply (4888 residents + 2200 tourists), the number of individuals affected by this outbreak was estimated to be 2600 (95% CI 2400–2900), assuming an exposure to tap water and a risk for gastroenteritis among the tourists similar to that among the inhabitants. Microbiological stool analyses Among the 35 stool samples analysed for bacteria, 11 (31.5%) were positive for C. coli. Twenty-four of the 35 samples were also tested for viruses: rotavirus A was detected in 17 (71.0%) and norovirus in five (21.0%) samples, while nine (37.5%) samples were co-infected (four C. coli + rotavirus; one C. coli + norovirus; three rotavirus + norovirus; and one C. coli + rotavirus + norovirus). Of the 11 C. coli isolates, six were characterised as biotype 2, and were identical following RAPD analysis, while the remaining five were of biotype 1. Rotavirus typing by RTPCR identified a single genotype P-(8), G1. Molecular characterisation of the norovirus isolates identified two different patterns, with three isolates being identical and close to the Saragota strain (genogroup I) and two close to the new GGIIb variant (genogroup II). Environmental investigation All of the water samples from source B were negative, regardless of the site of the distribution system tested. Analyses of the water originating from source A, sampled after leaving the water-works on 23 and 25 August, revealed high concentrations of E. coli (2263 / 100 mL), enterococci (415 / 100 mL), total coliforms (90 / 100 mL), faecal streptococci (136 / 100 mL) and sulphite-reducing clostridia (10 / 100 mL), and an absence of chlorination. Rotavirus A genotype G1 was detected in the crude water sampled from source A. The analyses for enterovirus, astrovirus, norovirus, hepatitis A virus and Campylobacter were negative. Source A was situated in a valley containing a river, below a hamlet of 15 dwellings, including farms breeding cattle, pork and sheep. The source was not supposed to be supplied by the river. The river resulted from the convergence of about ten different streams, with one originating close to a slaughterhouse, and was exposed to agricultural run-off and manure effluent; sheep and cattle grazed in meadows bordering the river, and a large amount of sewage from water treatment facilities was sprayed on to land close to the river. Cowsheds were close to the water catchment area. The immediate protected area around source A was too small and littered with building rubbish. Air vents without protection in the door made the source accessible to animals and insects. Bacteriological analyses of the water samples from source A and the river revealed that the contamination was of both human and animal origin, with a higher concentration in the river. The levels and trends of the chemical parameters of source A were similar to those of the river, strongly suggesting that the river was connected to source A (Fig. 3). A retrospective review of routine data collected between 1991 and 2000 showed irregularities in the maintenance of the water distribution system and the water treatment plant, particularly during the month of August. In the hamlet, the sewage system of the 15 households did not comply with sanitary rules. All but three ejected waste either into a cesspool or superficially without treatment. Three regions of gene 9 of the rotavirus detected in stools and water were compared. Sequencing showed that the strains were closely related but not identical. Assuming a likely onset of the outbreak between 14 and 19 August, and taking into account a minimum incubation period of 2 days for rotavirus A, norovirus and C. coli, the contamination of the water distribution system probably occurred between 12 and 17 August. Contamination probably persisted until 28 August, as intensive chlorination, beginning on 25 August, was effective only from 28 August onwards (Fig. 2). DISCUSSION The present descriptive and retrospective cohort studies confirmed that the outbreak of infection in the community of Gourdon was related to the consumption of tap water and affected 2600 individuals. The outbreak involved multiple pathogens, with group A rotavirus and C. coli being the most frequent, followed by norovirus. Results of environmental investigations suggested that water contamination was related to the contamination of the source, combined with a failure of the water chlorination system. The association with tap water consumption, the dose-effect relationship, and the fact that multiple pathogens were involved, were all strongly in favour of tap water as the origin of infection. Moreover, the number of cases decreased dramatically after the water restriction advice and the implementation of control measures on the water supply. No food products explored in the study were associated with illness. An AR of 14.5% and 31% was observed among adults and children aged < 6 years, respectively, who reported that they had not consumed tap water. This relatively high rate could be explained by several phenomena. Many individuals may have consumed tap water without being aware of it, e.g., in drinks or ice cubes consumed outside of the home, via raw fruits and vegetables, dishes prepared with unboiled water, or while cleaning their teeth. It can be hypothesised that people who became ill without being aware of having consumed tap water were exposed to small amounts of water and had been subjected to a small infective dose. In addition, children, more than adults, may have been contaminated through person-to-person transmission (e.g., family contact), which is the most common route of transmission for rotavirus and calicivirus. Waterborne contamination can affect large numbers of individuals [3,9–11]. In the present study, assuming a 50% increase in the population during the tourist season, the number of affected people was estimated to be between 2400 and 2900. The cohort study showed that almost half (46.5%) of the cases consulted a GP. Applying this proportion to the estimated number of cases yielded an estimated 1116–1349 ill individuals who consulted a GP for gastroenteritis, similar to the number of consultations for gastroenteritis actually reported by the GPs (n = 1037). The stool sample analyses revealed the presence of multiple pathogens (rotavirus A, norovirus genogroups I and II and C. coli) as single or co-infections. Except for C. coli, these pathogens have been implicated in numerous waterborne outbreaks, suggesting an influx of sewage into the water supply system [10,12–18]. Rotaviruses are known to be responsible for large epidemics during winter seasons, affecting particularly children aged < 3 years. These epidemics are caused by the faecal route of contamination (personto-person transmission). The high proportion of stools positive for rotavirus (71%) is uncommon in waterborne outbreaks and probably reflects massive contamination [15,16,19]. Different molecular sequences of rotavirus A genotype G1 found in stools and water indicated a human faecal source of infection. In the present study, contamination of water by sewage could be linked with the large amount of sewage from the water purification plant sprayed on to land close to the river and the water catchments. In this outbreak, C. coli, which was identified in 32% of stools, was the second most frequent pathogen. To our knowledge, human C. coli contamination has not been described previously as a causative agent in waterborne outbreaks, unlike C. jejuni [11,20–25], although C. coli has been isolated from water [26]. The main symptoms observed (diarrhoea, abdominal pain, nausea) and the mean duration of the disease (4.5 days) were compatible with a Campylobacter infection. No bloody diarrhoea was reported, although bloody diarrhoea is not constant in Campylobacter infections [27]. The high proportion of individuals who had to take sick leave, and the fact that six individuals were hospitalised, suggested a relative severity of symptoms, which is much more compatible with Campylobacter infection than with virus infections. Among the hospitalised patients, 60% were infected with C. coli. Among patients whose stools were analysed and whose age was known, cases infected only with rotavirus were younger (mean age 49 years; range, 5 months to 95 years) than cases infected only with Campylobacter (mean age 66 years; range 23–88 years) or co-infected with rotavirus plus Campylobacter (mean age 69 years; range 2–91 years). Moreover, the proportion (52%) of medical visits was unusually high compared with the proportion of visits observed during other waterborne outbreaks [13,16,28]. Campylobacter needs specific atmospheric conditions for growth and survival and its detection can be very difficult [29,30]. Because of a rapid physiological transformation to a viable, though non-cultivable, state, it has been suggested that conventional culture methods often fail [31]. Nevertheless, in the present study, the failure to detect C. coli in water, and the detection of noroviruses in stools, could be caused by the small (10 mL) water samples examined. Ha¨nninen et al. [30] proposed that the volume used for detection of suspected pathogens in drinking water should be at least 8–10 L. Groundwater is considered widely to be microbiologically clean, and is usually used for drinking without treatment. However, groundwater can be contaminated by the environment [26,32]. Stanley et al. [33] showed that Campylobacter can occur in groundwater, with hydrological evidence suggesting that the source of contamination in their study was a dairy farm situated within the hydrological catchment area of the groundwater, which was confirmed when identical strains of C. jejuni were isolated from groundwater and the dairy herd. In the present study, Campylobacter was not identified from groundwater, and dairy herds and other domestic animals in farms around the source were not investigated. However, results of bacterial, chemical and physical water analyses suggested strongly that a river contaminated by agricultural run-off from surrounding sheep and cattle grazing meadows was contaminating the source. In addition, the immediate area of the source was insufficiently protected and vulnerable to the environment. Wild birds, rodents, pigs and cows are all hosts for C. coli and could contaminate surface water easily [26,27,34,35]. Campylobacters are sensitive to many environmental factors and to chlorine [33,36,37], but a failure of the chlorination system could have resulted in the presence of a high concentration of C. coli. Waterborne outbreaks are often caused by several factors that act together [18,20,23,38]. In some outbreaks, heavy rainfall has been responsible for increasing agricultural run-off into rivers, although this did not occur before the outbreak in Gourdon. This underlies the need for increased frequency of routine water quality monitoring, chlorination of the water supply according to environmental conditions, and implementation of preventive measures. This outbreak raised the question of the delay in raising the alert. An outbreak can be recognised by: (i) the poor quality of water detected during regular controls; (ii) complaints about water quality; (iii) an incident in the supply and ℠or sewage water system during maintenance work; and (iv) an increased number of cases of gastroenteritis. In most instances, waterborne outbreaks are suspected when there are already a large number of declared cases [3,12,18,28,38,39]. In the present outbreak, the initial warning was given by a GP who notified officials 5–10 days after the assumed date of onset. This highlights the role of GPs in the detection of an unusual event, particularly during the summer, when an increase in acute gastroenteritis is less expected than during winter [40,41]. In addition, a routine water quality monitoring system able to detect contamination of drinking water before the occurrence of illness and complaints about the water quality should be used as an early warning source. The retrospective cohort study of Gourdon residents allowed the scale and severity of the outbreak to be quantified precisely and the hypothesis that tap water was the vehicle of the outbreak to be tested. When provided with a cohort sample representative of the study population, the number of people affected can be estimated. However, as the exposure of the population was high, a relatively large sample size was needed to test the association between the disease and the exposure. Because of misclassification of exposure (unrecognised consumption of tap water), the attributable risk is probably underestimated. In the absence of obvious alternative explanations, water contamination is the most likely common cause of an explosive outbreak of AGE occurring in a large population. Tillett et al. [42] proposed categorising levels of evidence for waterborne outbreaks. A descriptive study suggesting a water-related outbreak, together with the identification of the same pathogen in stools and water, should be sufficient to conclude that an outbreak is associated strongly with water consumption. Faecal indicators of water quality are currently insufficiently sensitive for some pathogens, notably viruses and parasites (e.g., Cryptosporidium). Therefore, in the absence of evidence from an analytical study to demonstrate an association, the link between illness and water can only be classified as probable or possible. However, total coliform counts are good indicators of faecal contamination with other pathogens (e.g., rotaviruses, E. coli, Campylobacter) [43]. In conclusion, the waterborne outbreak described in the present study was most probably caused by faecal contamination of source A by groundwater, resulting from deficiencies in the maintenance of the water distribution system. Corrective measures were implemented based on the results of the investigation. To our knowledge, this is the first documented waterborne outbreak involving human C. coli infection. In-depth environmental investigations to identify the origin of water contamination are important in order to better understand microbial transmission from the environment and to implement appropriate controls and preventive measures. In addition, surveys on vehicles and vectors that transmit Campylobacter between hosts are crucial in order to better understand the epidemiology of these organisms.

What was the event?

278

A large multi-pathogen waterborne community outbreak linked to faecal contamination of a groundwater system, France, 2000

A large waterborne outbreak of infection that occurred during August 2000 in a local community in France was investigated initially via a rapid survey of visits to local physicians. A retrospective cohort study was then conducted on a random cluster sample of residents. Of 709 residents interviewed, 202 (28.5%) were definite cases (at least three liquid stools ℠day or vomiting) and 62 (8.7%) were probable cases (less than three liquid stools ℠day or abdominal pain). Those who had drunk tap water had a threefold increased risk for illness (95% CI 2.4–4.0). The risk increased with the amount of water consumed (chi-square trend: p < 0.0001). Bacteriological analyses of stools were performed for 35 patients and virological analyses for 24 patients. Campylobacter coli, group A rotavirus and norovirus were detected in 31.5%, 71.0% and 21% of samples, respectively. An extensive environmental investigation concluded that a groundwater source to this community had probably been contaminated by agricultural run-off, and a failure in the chlorination system was identified. This is the first documented waterborne outbreak of infection involving human C. coli infections. A better understanding of the factors influencing campylobacter transmission between hosts is required. Keywords Campylobacter coli, epidemiology, gastroenteritis, norovirus, rotavirus, waterborne outbreak INTRODUCTION Contaminated drinking water causes extensive outbreaks of illness because of the large number of people served by water supply facilities. Detection of such outbreaks requires the identification of an increase in illness (usually the rate of gastrointestinal disease) in the exposed population and confirmation that water was the route of transmission [1]. Many countries have routine reporting systems for detecting foodborne and waterborne outbreaks of infection. However, in routine practice it is not easy to detect an increase in clinical cases and to link this increase to waterborne transmission; thus waterborne outbreaks are often unrecognised and underestimated [2]. Microbiological examination of water is more complex than examination of stools, and there is often a failure to detect pathogens in water. Norovirus, Cryptosporidium parvum, Giardia intestinalis and Campylobacter jejuni are the pathogens identified most frequently [3]. The present study describes a large waterborne community outbreak of infection with multiple pathogens, including Campylobacter coli. METHODS AND MATERIALS Background On 23 August 2000, a general practitioner (GP) informed the local district health department that 16 cases of acute gastroenteritis (AGE) had occurred among residents of a holiday camp in the Gourdon community since 21 August. Gourdon, which is located in the south-west of France, has a population of 4888 inhabitants, which increases by 50% in the July– August tourist season. Investigations at the holiday camp concluded that foodborne transmission was very unlikely, but found that tap water was highly contaminated with faecal coliforms. Preliminary information collected from physicians in Gourdon indicated an increase of the number of consultations for AGE. Gourdon tap water is supplied by two underground water facilities: facility A serving 3800 water consumers in Gourdon; and facility B serving 1200 water consumers in Gourdon and the surrounding communities. Water treatment in facility A consists of chlorination, which involves two automatic pumps working in rotation. The water then passes through two semi in-ground reservoir tanks (1100 m3 and 200 m3, respectively) before distribution. Facility B water is mixed with a fraction of facility A water in another 500 m3 semi in-ground reservoir tank before being distributed. This mixing with the treated water from facility A is intended to disinfect the water from facility B. Since the increase in AGE was noted only in Gourdon, contaminated tap water from facility A was suspected to be the origin of the outbreak. Therefore, the population of Gourdon was informed on 25 August to avoid the consumption of tap water or to boil it for 5 min. Survey of gastroenteritis-related medical consultations To confirm the outbreak, the 11 GPs in Gourdon and the local hospital emergency department were requested to provide the number of total daily visits and the number of cases of gastroenteritis cases occurring between 1 August and 11 September 2000. Retrospective cohort study A retrospective cohort study was performed on a sample of Gourdon households between 5 and 15 September 2000. Domestic households were selected randomly from the telephone directory. A definite case was defined as a Gourdon resident who had at least three liquid stools ℠day or vomiting between 1 August and 3 September, and who had been present in Gourdon for ‡ 1 day between 1 and 31 August; if less than three liquid stools ℠day or abdominal pain had occurred, the case was classified as probable. Exposure was defined as the consumption of tap water before 25 August 2000. A standardised telephone questionnaire was completed to obtain demographic data (gender, date of birth), food consumption (commercially prepared dishes, sea-food, raw milk, pastry), tap water consumption (before and after 25 August, daily number of glasses of water, consumption of boiled tap water), symptoms of gastroenteritis (date of onset, number of liquid stools ℠day, vomiting, nausea, abdominal pain, fever), medical care (in- or out-patient) and consequences on daily activities (sick leave, bed confinement) for each member of the selected household. Under the hypothesis that 60% of the population consumed tap water, an estimated 550 people were required for inclusion in the study to identify a relative risk (RR) of ‡ 3, with a 5% a risk and a power of 80%. Based on an average of two individuals per household (INSERM 1999 population census), 300 households were required for inclusion in the survey. The association between AGE and the consumption of tap water was assessed by calculating the risk ratio (RR) and 95% CI, taking into account a household cluster design effect. A dose-effect relationship was evaluated by analysing the RR trend for different categories of daily number of glasses of tap water consumed (chi-square trend). Proportions were compared using the corrected chi-square test, and means were compared using the Student test or Mann and Witney test. Data were analysed with Epi-Info 6.04c (CDC, Atlanta, GA, USA). The gender and age distribution of the cohort population were compared with that of the general population of Gourdon (INSERM 1999 population census) to assess whether it was representative. Under the hypothesis of representation, the total number of cases of gastroenteritis was estimated by applying the attack rate to the total population of Gourdon, including the tourist population (data provided by the Gourdon tourism office). The most likely period for contamination of the water distribution system was based on the epidemic curve and the minimal incubation period of the pathogens involved. Microbiological analyses of stools Stool samples were examined for the presence of Salmonella spp., Shigella spp., Staphylococcus aureus, Campylobacter spp., Yersinia enterocolitica, Escherichia coli, enterococci, rotavirus groups A and C, astrovirus, calicivirus, adenovirus types 40 and 41, enterovirus, hepatitis A virus and Cryptosporidum. Viruses were detected using an enzyme immunoassay for group A rotavirus, astrovirus and adenovirus 40 ℠41, and by RT-PCR for calicivirus, group C rotavirus, enterovirus and hepatitis A virus [4–6]. Campylobacter isolates were sent to the National Reference Centre for Campylobacter for identification and fingerprinting. Species identification was by means of phenotypic tests (API Campy; bioMe´rieux, Lyon, France), with use of PCR to differentiate Campylobacter jejuni from Camplylobacter coli [7]. Random amplified polymorphic DNA fingerprinting was performed with primer 3881, 5¢-AACGCGCAAC. Environmental investigation An environmental and microbiological investigation of the water distribution system was performed. The inspection concerned the water catchments, the water distribution and sewage system, agricultural practices and private households around the sources. Bacteriological (Escherichia coli, enterococci, faecal streptococci, thermotolerant coliforms, sulphite-reducing clostridia) and chemical (chlorine) analyses were performed on tap water and on the water sampled from the two groundwater sources. After 23 August, the water chlorine concentration and the presence of coliform bacteria were monitored daily at several sites of the water system before and after the resort treatment. Examinations for viruses (group A rotavirus, astrovirus, norovirus, enterovirus, hepatitis A virus) and Campylobacter spp. were undertaken on two 10-mL tap water samples collected inhouse on 23 August, and on two 10-mL samples from groundwater sources A and B on 25 August. Viruses were detected by RT-PCR, followed by hybridisation with a specific primer to improve sensitivity, while C. jejuni and C. coli were detected by PCR with specific primers. Group A rotavirus isolates detected in stools and in water were sequenced using Beg 9 and End 9 primers [8] and compared using the Infobiogene GCG software (http://www.infobiogen.fr). Bacteriological (revived 36C, revived 22C, thermotolerant coliforms, streptococci and faecal enterococci), physical (temperature, conductivity) and chemical (oxygen consumption, ammonium, nitrates, phosphorus) parameters from source A and the river were analysed. Trends were compared for an 8-week period between 21 December 2000 and 7 February 2001. RESULTS Epidemiology Between 1 August and 11 September 2000, 7104 Gourdon residents consulted a GP or the local hospital emergency department. Between 24 and 28 August, gastroenteritis accounted for 44% (479 ℠1097) of all medical visits, compared with 6% (126 ℠2185) during the first 15 days of August (Fig. 1). Among the 498 households contacted for the household survey, 198 were excluded (103 did not respond, 49 did not fulfil the inclusion criteria, 43 refused to participate, and three for unknown reasons). The questionnaire was completed for 300 households (709 individuals). Data were missing concerning gender for one individual and age for 11 individuals. Overall, 331 (46.7%; 95% CI 44.5–49.0) of respondents were male and 377 (53.2%; 95% CI 50.9–55.5) were female. The median age was 44 years (range < 1–94 years). The age and gender distribution of the sample was similar to that of the general population. During the outbreak period, 264 of the 709 respondents (attack rate (AR) 37.2%) had been ill, with 202 definite cases (AR 28.5%; 95% CI 24.6– 32.4) and 62 probable cases (AR 8.7%; 95% CI 6.5– 11.0). The AR was higher in females (31.8%) than in males (24.8%), and was highest in children aged < 6 years (42.1%, Table 1). The epidemic curve (Fig. 2) suggests that the outbreak started between 14 and 19 August. The shape of the epidemic curve in the retrospective cohort study was very similar to that based on the number of visits to GPs for gastroenteritis. Patients complained mainly of diarrhoea, abdominal pain and nausea (Table 2). No cases of bloody diarrhoea were notified. The mean duration of symptoms was 4.5 days (median 3, range 1–30 days). Among definite cases, 52% had visited a GP, 42% had taken sick leave for ‡ 1 day, 32% had required bed rest, and 2.9% (six patients) had been hospitalised. In total, 336 (47.4%) respondents had drunk tap water before 25 August, whereas 646 (91.1%) did not drink tap water after 25 August. The risk for illness, taking into account the sample cluster design effect of 1.4, was three-fold greater (95% CI 2.2–4.1) for those who had drunk tap water before 25 August, and increased with the amount of water consumed (chi-square trend: p < 0.0001; Table 3). The risk of gastroenteritis associated with drinking tap water did not change substantially after adjustment for age (RRMH 3.3, 95% CI 2.5–4.4), and there was no modification effect according to age. Food consumption was not associated with illness. Applying the AR of 37% to the total population of Gourdon exposed to the water supply (4888 residents + 2200 tourists), the number of individuals affected by this outbreak was estimated to be 2600 (95% CI 2400–2900), assuming an exposure to tap water and a risk for gastroenteritis among the tourists similar to that among the inhabitants. Microbiological stool analyses Among the 35 stool samples analysed for bacteria, 11 (31.5%) were positive for C. coli. Twenty-four of the 35 samples were also tested for viruses: rotavirus A was detected in 17 (71.0%) and norovirus in five (21.0%) samples, while nine (37.5%) samples were co-infected (four C. coli + rotavirus; one C. coli + norovirus; three rotavirus + norovirus; and one C. coli + rotavirus + norovirus). Of the 11 C. coli isolates, six were characterised as biotype 2, and were identical following RAPD analysis, while the remaining five were of biotype 1. Rotavirus typing by RTPCR identified a single genotype P-(8), G1. Molecular characterisation of the norovirus isolates identified two different patterns, with three isolates being identical and close to the Saragota strain (genogroup I) and two close to the new GGIIb variant (genogroup II). Environmental investigation All of the water samples from source B were negative, regardless of the site of the distribution system tested. Analyses of the water originating from source A, sampled after leaving the water-works on 23 and 25 August, revealed high concentrations of E. coli (2263 / 100 mL), enterococci (415 / 100 mL), total coliforms (90 / 100 mL), faecal streptococci (136 / 100 mL) and sulphite-reducing clostridia (10 / 100 mL), and an absence of chlorination. Rotavirus A genotype G1 was detected in the crude water sampled from source A. The analyses for enterovirus, astrovirus, norovirus, hepatitis A virus and Campylobacter were negative. Source A was situated in a valley containing a river, below a hamlet of 15 dwellings, including farms breeding cattle, pork and sheep. The source was not supposed to be supplied by the river. The river resulted from the convergence of about ten different streams, with one originating close to a slaughterhouse, and was exposed to agricultural run-off and manure effluent; sheep and cattle grazed in meadows bordering the river, and a large amount of sewage from water treatment facilities was sprayed on to land close to the river. Cowsheds were close to the water catchment area. The immediate protected area around source A was too small and littered with building rubbish. Air vents without protection in the door made the source accessible to animals and insects. Bacteriological analyses of the water samples from source A and the river revealed that the contamination was of both human and animal origin, with a higher concentration in the river. The levels and trends of the chemical parameters of source A were similar to those of the river, strongly suggesting that the river was connected to source A (Fig. 3). A retrospective review of routine data collected between 1991 and 2000 showed irregularities in the maintenance of the water distribution system and the water treatment plant, particularly during the month of August. In the hamlet, the sewage system of the 15 households did not comply with sanitary rules. All but three ejected waste either into a cesspool or superficially without treatment. Three regions of gene 9 of the rotavirus detected in stools and water were compared. Sequencing showed that the strains were closely related but not identical. Assuming a likely onset of the outbreak between 14 and 19 August, and taking into account a minimum incubation period of 2 days for rotavirus A, norovirus and C. coli, the contamination of the water distribution system probably occurred between 12 and 17 August. Contamination probably persisted until 28 August, as intensive chlorination, beginning on 25 August, was effective only from 28 August onwards (Fig. 2). DISCUSSION The present descriptive and retrospective cohort studies confirmed that the outbreak of infection in the community of Gourdon was related to the consumption of tap water and affected 2600 individuals. The outbreak involved multiple pathogens, with group A rotavirus and C. coli being the most frequent, followed by norovirus. Results of environmental investigations suggested that water contamination was related to the contamination of the source, combined with a failure of the water chlorination system. The association with tap water consumption, the dose-effect relationship, and the fact that multiple pathogens were involved, were all strongly in favour of tap water as the origin of infection. Moreover, the number of cases decreased dramatically after the water restriction advice and the implementation of control measures on the water supply. No food products explored in the study were associated with illness. An AR of 14.5% and 31% was observed among adults and children aged < 6 years, respectively, who reported that they had not consumed tap water. This relatively high rate could be explained by several phenomena. Many individuals may have consumed tap water without being aware of it, e.g., in drinks or ice cubes consumed outside of the home, via raw fruits and vegetables, dishes prepared with unboiled water, or while cleaning their teeth. It can be hypothesised that people who became ill without being aware of having consumed tap water were exposed to small amounts of water and had been subjected to a small infective dose. In addition, children, more than adults, may have been contaminated through person-to-person transmission (e.g., family contact), which is the most common route of transmission for rotavirus and calicivirus. Waterborne contamination can affect large numbers of individuals [3,9–11]. In the present study, assuming a 50% increase in the population during the tourist season, the number of affected people was estimated to be between 2400 and 2900. The cohort study showed that almost half (46.5%) of the cases consulted a GP. Applying this proportion to the estimated number of cases yielded an estimated 1116–1349 ill individuals who consulted a GP for gastroenteritis, similar to the number of consultations for gastroenteritis actually reported by the GPs (n = 1037). The stool sample analyses revealed the presence of multiple pathogens (rotavirus A, norovirus genogroups I and II and C. coli) as single or co-infections. Except for C. coli, these pathogens have been implicated in numerous waterborne outbreaks, suggesting an influx of sewage into the water supply system [10,12–18]. Rotaviruses are known to be responsible for large epidemics during winter seasons, affecting particularly children aged < 3 years. These epidemics are caused by the faecal route of contamination (personto-person transmission). The high proportion of stools positive for rotavirus (71%) is uncommon in waterborne outbreaks and probably reflects massive contamination [15,16,19]. Different molecular sequences of rotavirus A genotype G1 found in stools and water indicated a human faecal source of infection. In the present study, contamination of water by sewage could be linked with the large amount of sewage from the water purification plant sprayed on to land close to the river and the water catchments. In this outbreak, C. coli, which was identified in 32% of stools, was the second most frequent pathogen. To our knowledge, human C. coli contamination has not been described previously as a causative agent in waterborne outbreaks, unlike C. jejuni [11,20–25], although C. coli has been isolated from water [26]. The main symptoms observed (diarrhoea, abdominal pain, nausea) and the mean duration of the disease (4.5 days) were compatible with a Campylobacter infection. No bloody diarrhoea was reported, although bloody diarrhoea is not constant in Campylobacter infections [27]. The high proportion of individuals who had to take sick leave, and the fact that six individuals were hospitalised, suggested a relative severity of symptoms, which is much more compatible with Campylobacter infection than with virus infections. Among the hospitalised patients, 60% were infected with C. coli. Among patients whose stools were analysed and whose age was known, cases infected only with rotavirus were younger (mean age 49 years; range, 5 months to 95 years) than cases infected only with Campylobacter (mean age 66 years; range 23–88 years) or co-infected with rotavirus plus Campylobacter (mean age 69 years; range 2–91 years). Moreover, the proportion (52%) of medical visits was unusually high compared with the proportion of visits observed during other waterborne outbreaks [13,16,28]. Campylobacter needs specific atmospheric conditions for growth and survival and its detection can be very difficult [29,30]. Because of a rapid physiological transformation to a viable, though non-cultivable, state, it has been suggested that conventional culture methods often fail [31]. Nevertheless, in the present study, the failure to detect C. coli in water, and the detection of noroviruses in stools, could be caused by the small (10 mL) water samples examined. Ha¨nninen et al. [30] proposed that the volume used for detection of suspected pathogens in drinking water should be at least 8–10 L. Groundwater is considered widely to be microbiologically clean, and is usually used for drinking without treatment. However, groundwater can be contaminated by the environment [26,32]. Stanley et al. [33] showed that Campylobacter can occur in groundwater, with hydrological evidence suggesting that the source of contamination in their study was a dairy farm situated within the hydrological catchment area of the groundwater, which was confirmed when identical strains of C. jejuni were isolated from groundwater and the dairy herd. In the present study, Campylobacter was not identified from groundwater, and dairy herds and other domestic animals in farms around the source were not investigated. However, results of bacterial, chemical and physical water analyses suggested strongly that a river contaminated by agricultural run-off from surrounding sheep and cattle grazing meadows was contaminating the source. In addition, the immediate area of the source was insufficiently protected and vulnerable to the environment. Wild birds, rodents, pigs and cows are all hosts for C. coli and could contaminate surface water easily [26,27,34,35]. Campylobacters are sensitive to many environmental factors and to chlorine [33,36,37], but a failure of the chlorination system could have resulted in the presence of a high concentration of C. coli. Waterborne outbreaks are often caused by several factors that act together [18,20,23,38]. In some outbreaks, heavy rainfall has been responsible for increasing agricultural run-off into rivers, although this did not occur before the outbreak in Gourdon. This underlies the need for increased frequency of routine water quality monitoring, chlorination of the water supply according to environmental conditions, and implementation of preventive measures. This outbreak raised the question of the delay in raising the alert. An outbreak can be recognised by: (i) the poor quality of water detected during regular controls; (ii) complaints about water quality; (iii) an incident in the supply and ℠or sewage water system during maintenance work; and (iv) an increased number of cases of gastroenteritis. In most instances, waterborne outbreaks are suspected when there are already a large number of declared cases [3,12,18,28,38,39]. In the present outbreak, the initial warning was given by a GP who notified officials 5–10 days after the assumed date of onset. This highlights the role of GPs in the detection of an unusual event, particularly during the summer, when an increase in acute gastroenteritis is less expected than during winter [40,41]. In addition, a routine water quality monitoring system able to detect contamination of drinking water before the occurrence of illness and complaints about the water quality should be used as an early warning source. The retrospective cohort study of Gourdon residents allowed the scale and severity of the outbreak to be quantified precisely and the hypothesis that tap water was the vehicle of the outbreak to be tested. When provided with a cohort sample representative of the study population, the number of people affected can be estimated. However, as the exposure of the population was high, a relatively large sample size was needed to test the association between the disease and the exposure. Because of misclassification of exposure (unrecognised consumption of tap water), the attributable risk is probably underestimated. In the absence of obvious alternative explanations, water contamination is the most likely common cause of an explosive outbreak of AGE occurring in a large population. Tillett et al. [42] proposed categorising levels of evidence for waterborne outbreaks. A descriptive study suggesting a water-related outbreak, together with the identification of the same pathogen in stools and water, should be sufficient to conclude that an outbreak is associated strongly with water consumption. Faecal indicators of water quality are currently insufficiently sensitive for some pathogens, notably viruses and parasites (e.g., Cryptosporidium). Therefore, in the absence of evidence from an analytical study to demonstrate an association, the link between illness and water can only be classified as probable or possible. However, total coliform counts are good indicators of faecal contamination with other pathogens (e.g., rotaviruses, E. coli, Campylobacter) [43]. In conclusion, the waterborne outbreak described in the present study was most probably caused by faecal contamination of source A by groundwater, resulting from deficiencies in the maintenance of the water distribution system. Corrective measures were implemented based on the results of the investigation. To our knowledge, this is the first documented waterborne outbreak involving human C. coli infection. In-depth environmental investigations to identify the origin of water contamination are important in order to better understand microbial transmission from the environment and to implement appropriate controls and preventive measures. In addition, surveys on vehicles and vectors that transmit Campylobacter between hosts are crucial in order to better understand the epidemiology of these organisms.

When did this happen?

279

A large multi-pathogen waterborne community outbreak linked to faecal contamination of a groundwater system, France, 2000

A large waterborne outbreak of infection that occurred during August 2000 in a local community in France was investigated initially via a rapid survey of visits to local physicians. A retrospective cohort study was then conducted on a random cluster sample of residents. Of 709 residents interviewed, 202 (28.5%) were definite cases (at least three liquid stools ℠day or vomiting) and 62 (8.7%) were probable cases (less than three liquid stools ℠day or abdominal pain). Those who had drunk tap water had a threefold increased risk for illness (95% CI 2.4–4.0). The risk increased with the amount of water consumed (chi-square trend: p < 0.0001). Bacteriological analyses of stools were performed for 35 patients and virological analyses for 24 patients. Campylobacter coli, group A rotavirus and norovirus were detected in 31.5%, 71.0% and 21% of samples, respectively. An extensive environmental investigation concluded that a groundwater source to this community had probably been contaminated by agricultural run-off, and a failure in the chlorination system was identified. This is the first documented waterborne outbreak of infection involving human C. coli infections. A better understanding of the factors influencing campylobacter transmission between hosts is required. Keywords Campylobacter coli, epidemiology, gastroenteritis, norovirus, rotavirus, waterborne outbreak INTRODUCTION Contaminated drinking water causes extensive outbreaks of illness because of the large number of people served by water supply facilities. Detection of such outbreaks requires the identification of an increase in illness (usually the rate of gastrointestinal disease) in the exposed population and confirmation that water was the route of transmission [1]. Many countries have routine reporting systems for detecting foodborne and waterborne outbreaks of infection. However, in routine practice it is not easy to detect an increase in clinical cases and to link this increase to waterborne transmission; thus waterborne outbreaks are often unrecognised and underestimated [2]. Microbiological examination of water is more complex than examination of stools, and there is often a failure to detect pathogens in water. Norovirus, Cryptosporidium parvum, Giardia intestinalis and Campylobacter jejuni are the pathogens identified most frequently [3]. The present study describes a large waterborne community outbreak of infection with multiple pathogens, including Campylobacter coli. METHODS AND MATERIALS Background On 23 August 2000, a general practitioner (GP) informed the local district health department that 16 cases of acute gastroenteritis (AGE) had occurred among residents of a holiday camp in the Gourdon community since 21 August. Gourdon, which is located in the south-west of France, has a population of 4888 inhabitants, which increases by 50% in the July– August tourist season. Investigations at the holiday camp concluded that foodborne transmission was very unlikely, but found that tap water was highly contaminated with faecal coliforms. Preliminary information collected from physicians in Gourdon indicated an increase of the number of consultations for AGE. Gourdon tap water is supplied by two underground water facilities: facility A serving 3800 water consumers in Gourdon; and facility B serving 1200 water consumers in Gourdon and the surrounding communities. Water treatment in facility A consists of chlorination, which involves two automatic pumps working in rotation. The water then passes through two semi in-ground reservoir tanks (1100 m3 and 200 m3, respectively) before distribution. Facility B water is mixed with a fraction of facility A water in another 500 m3 semi in-ground reservoir tank before being distributed. This mixing with the treated water from facility A is intended to disinfect the water from facility B. Since the increase in AGE was noted only in Gourdon, contaminated tap water from facility A was suspected to be the origin of the outbreak. Therefore, the population of Gourdon was informed on 25 August to avoid the consumption of tap water or to boil it for 5 min. Survey of gastroenteritis-related medical consultations To confirm the outbreak, the 11 GPs in Gourdon and the local hospital emergency department were requested to provide the number of total daily visits and the number of cases of gastroenteritis cases occurring between 1 August and 11 September 2000. Retrospective cohort study A retrospective cohort study was performed on a sample of Gourdon households between 5 and 15 September 2000. Domestic households were selected randomly from the telephone directory. A definite case was defined as a Gourdon resident who had at least three liquid stools ℠day or vomiting between 1 August and 3 September, and who had been present in Gourdon for ‡ 1 day between 1 and 31 August; if less than three liquid stools ℠day or abdominal pain had occurred, the case was classified as probable. Exposure was defined as the consumption of tap water before 25 August 2000. A standardised telephone questionnaire was completed to obtain demographic data (gender, date of birth), food consumption (commercially prepared dishes, sea-food, raw milk, pastry), tap water consumption (before and after 25 August, daily number of glasses of water, consumption of boiled tap water), symptoms of gastroenteritis (date of onset, number of liquid stools ℠day, vomiting, nausea, abdominal pain, fever), medical care (in- or out-patient) and consequences on daily activities (sick leave, bed confinement) for each member of the selected household. Under the hypothesis that 60% of the population consumed tap water, an estimated 550 people were required for inclusion in the study to identify a relative risk (RR) of ‡ 3, with a 5% a risk and a power of 80%. Based on an average of two individuals per household (INSERM 1999 population census), 300 households were required for inclusion in the survey. The association between AGE and the consumption of tap water was assessed by calculating the risk ratio (RR) and 95% CI, taking into account a household cluster design effect. A dose-effect relationship was evaluated by analysing the RR trend for different categories of daily number of glasses of tap water consumed (chi-square trend). Proportions were compared using the corrected chi-square test, and means were compared using the Student test or Mann and Witney test. Data were analysed with Epi-Info 6.04c (CDC, Atlanta, GA, USA). The gender and age distribution of the cohort population were compared with that of the general population of Gourdon (INSERM 1999 population census) to assess whether it was representative. Under the hypothesis of representation, the total number of cases of gastroenteritis was estimated by applying the attack rate to the total population of Gourdon, including the tourist population (data provided by the Gourdon tourism office). The most likely period for contamination of the water distribution system was based on the epidemic curve and the minimal incubation period of the pathogens involved. Microbiological analyses of stools Stool samples were examined for the presence of Salmonella spp., Shigella spp., Staphylococcus aureus, Campylobacter spp., Yersinia enterocolitica, Escherichia coli, enterococci, rotavirus groups A and C, astrovirus, calicivirus, adenovirus types 40 and 41, enterovirus, hepatitis A virus and Cryptosporidum. Viruses were detected using an enzyme immunoassay for group A rotavirus, astrovirus and adenovirus 40 ℠41, and by RT-PCR for calicivirus, group C rotavirus, enterovirus and hepatitis A virus [4–6]. Campylobacter isolates were sent to the National Reference Centre for Campylobacter for identification and fingerprinting. Species identification was by means of phenotypic tests (API Campy; bioMe´rieux, Lyon, France), with use of PCR to differentiate Campylobacter jejuni from Camplylobacter coli [7]. Random amplified polymorphic DNA fingerprinting was performed with primer 3881, 5¢-AACGCGCAAC. Environmental investigation An environmental and microbiological investigation of the water distribution system was performed. The inspection concerned the water catchments, the water distribution and sewage system, agricultural practices and private households around the sources. Bacteriological (Escherichia coli, enterococci, faecal streptococci, thermotolerant coliforms, sulphite-reducing clostridia) and chemical (chlorine) analyses were performed on tap water and on the water sampled from the two groundwater sources. After 23 August, the water chlorine concentration and the presence of coliform bacteria were monitored daily at several sites of the water system before and after the resort treatment. Examinations for viruses (group A rotavirus, astrovirus, norovirus, enterovirus, hepatitis A virus) and Campylobacter spp. were undertaken on two 10-mL tap water samples collected inhouse on 23 August, and on two 10-mL samples from groundwater sources A and B on 25 August. Viruses were detected by RT-PCR, followed by hybridisation with a specific primer to improve sensitivity, while C. jejuni and C. coli were detected by PCR with specific primers. Group A rotavirus isolates detected in stools and in water were sequenced using Beg 9 and End 9 primers [8] and compared using the Infobiogene GCG software (http://www.infobiogen.fr). Bacteriological (revived 36C, revived 22C, thermotolerant coliforms, streptococci and faecal enterococci), physical (temperature, conductivity) and chemical (oxygen consumption, ammonium, nitrates, phosphorus) parameters from source A and the river were analysed. Trends were compared for an 8-week period between 21 December 2000 and 7 February 2001. RESULTS Epidemiology Between 1 August and 11 September 2000, 7104 Gourdon residents consulted a GP or the local hospital emergency department. Between 24 and 28 August, gastroenteritis accounted for 44% (479 ℠1097) of all medical visits, compared with 6% (126 ℠2185) during the first 15 days of August (Fig. 1). Among the 498 households contacted for the household survey, 198 were excluded (103 did not respond, 49 did not fulfil the inclusion criteria, 43 refused to participate, and three for unknown reasons). The questionnaire was completed for 300 households (709 individuals). Data were missing concerning gender for one individual and age for 11 individuals. Overall, 331 (46.7%; 95% CI 44.5–49.0) of respondents were male and 377 (53.2%; 95% CI 50.9–55.5) were female. The median age was 44 years (range < 1–94 years). The age and gender distribution of the sample was similar to that of the general population. During the outbreak period, 264 of the 709 respondents (attack rate (AR) 37.2%) had been ill, with 202 definite cases (AR 28.5%; 95% CI 24.6– 32.4) and 62 probable cases (AR 8.7%; 95% CI 6.5– 11.0). The AR was higher in females (31.8%) than in males (24.8%), and was highest in children aged < 6 years (42.1%, Table 1). The epidemic curve (Fig. 2) suggests that the outbreak started between 14 and 19 August. The shape of the epidemic curve in the retrospective cohort study was very similar to that based on the number of visits to GPs for gastroenteritis. Patients complained mainly of diarrhoea, abdominal pain and nausea (Table 2). No cases of bloody diarrhoea were notified. The mean duration of symptoms was 4.5 days (median 3, range 1–30 days). Among definite cases, 52% had visited a GP, 42% had taken sick leave for ‡ 1 day, 32% had required bed rest, and 2.9% (six patients) had been hospitalised. In total, 336 (47.4%) respondents had drunk tap water before 25 August, whereas 646 (91.1%) did not drink tap water after 25 August. The risk for illness, taking into account the sample cluster design effect of 1.4, was three-fold greater (95% CI 2.2–4.1) for those who had drunk tap water before 25 August, and increased with the amount of water consumed (chi-square trend: p < 0.0001; Table 3). The risk of gastroenteritis associated with drinking tap water did not change substantially after adjustment for age (RRMH 3.3, 95% CI 2.5–4.4), and there was no modification effect according to age. Food consumption was not associated with illness. Applying the AR of 37% to the total population of Gourdon exposed to the water supply (4888 residents + 2200 tourists), the number of individuals affected by this outbreak was estimated to be 2600 (95% CI 2400–2900), assuming an exposure to tap water and a risk for gastroenteritis among the tourists similar to that among the inhabitants. Microbiological stool analyses Among the 35 stool samples analysed for bacteria, 11 (31.5%) were positive for C. coli. Twenty-four of the 35 samples were also tested for viruses: rotavirus A was detected in 17 (71.0%) and norovirus in five (21.0%) samples, while nine (37.5%) samples were co-infected (four C. coli + rotavirus; one C. coli + norovirus; three rotavirus + norovirus; and one C. coli + rotavirus + norovirus). Of the 11 C. coli isolates, six were characterised as biotype 2, and were identical following RAPD analysis, while the remaining five were of biotype 1. Rotavirus typing by RTPCR identified a single genotype P-(8), G1. Molecular characterisation of the norovirus isolates identified two different patterns, with three isolates being identical and close to the Saragota strain (genogroup I) and two close to the new GGIIb variant (genogroup II). Environmental investigation All of the water samples from source B were negative, regardless of the site of the distribution system tested. Analyses of the water originating from source A, sampled after leaving the water-works on 23 and 25 August, revealed high concentrations of E. coli (2263 / 100 mL), enterococci (415 / 100 mL), total coliforms (90 / 100 mL), faecal streptococci (136 / 100 mL) and sulphite-reducing clostridia (10 / 100 mL), and an absence of chlorination. Rotavirus A genotype G1 was detected in the crude water sampled from source A. The analyses for enterovirus, astrovirus, norovirus, hepatitis A virus and Campylobacter were negative. Source A was situated in a valley containing a river, below a hamlet of 15 dwellings, including farms breeding cattle, pork and sheep. The source was not supposed to be supplied by the river. The river resulted from the convergence of about ten different streams, with one originating close to a slaughterhouse, and was exposed to agricultural run-off and manure effluent; sheep and cattle grazed in meadows bordering the river, and a large amount of sewage from water treatment facilities was sprayed on to land close to the river. Cowsheds were close to the water catchment area. The immediate protected area around source A was too small and littered with building rubbish. Air vents without protection in the door made the source accessible to animals and insects. Bacteriological analyses of the water samples from source A and the river revealed that the contamination was of both human and animal origin, with a higher concentration in the river. The levels and trends of the chemical parameters of source A were similar to those of the river, strongly suggesting that the river was connected to source A (Fig. 3). A retrospective review of routine data collected between 1991 and 2000 showed irregularities in the maintenance of the water distribution system and the water treatment plant, particularly during the month of August. In the hamlet, the sewage system of the 15 households did not comply with sanitary rules. All but three ejected waste either into a cesspool or superficially without treatment. Three regions of gene 9 of the rotavirus detected in stools and water were compared. Sequencing showed that the strains were closely related but not identical. Assuming a likely onset of the outbreak between 14 and 19 August, and taking into account a minimum incubation period of 2 days for rotavirus A, norovirus and C. coli, the contamination of the water distribution system probably occurred between 12 and 17 August. Contamination probably persisted until 28 August, as intensive chlorination, beginning on 25 August, was effective only from 28 August onwards (Fig. 2). DISCUSSION The present descriptive and retrospective cohort studies confirmed that the outbreak of infection in the community of Gourdon was related to the consumption of tap water and affected 2600 individuals. The outbreak involved multiple pathogens, with group A rotavirus and C. coli being the most frequent, followed by norovirus. Results of environmental investigations suggested that water contamination was related to the contamination of the source, combined with a failure of the water chlorination system. The association with tap water consumption, the dose-effect relationship, and the fact that multiple pathogens were involved, were all strongly in favour of tap water as the origin of infection. Moreover, the number of cases decreased dramatically after the water restriction advice and the implementation of control measures on the water supply. No food products explored in the study were associated with illness. An AR of 14.5% and 31% was observed among adults and children aged < 6 years, respectively, who reported that they had not consumed tap water. This relatively high rate could be explained by several phenomena. Many individuals may have consumed tap water without being aware of it, e.g., in drinks or ice cubes consumed outside of the home, via raw fruits and vegetables, dishes prepared with unboiled water, or while cleaning their teeth. It can be hypothesised that people who became ill without being aware of having consumed tap water were exposed to small amounts of water and had been subjected to a small infective dose. In addition, children, more than adults, may have been contaminated through person-to-person transmission (e.g., family contact), which is the most common route of transmission for rotavirus and calicivirus. Waterborne contamination can affect large numbers of individuals [3,9–11]. In the present study, assuming a 50% increase in the population during the tourist season, the number of affected people was estimated to be between 2400 and 2900. The cohort study showed that almost half (46.5%) of the cases consulted a GP. Applying this proportion to the estimated number of cases yielded an estimated 1116–1349 ill individuals who consulted a GP for gastroenteritis, similar to the number of consultations for gastroenteritis actually reported by the GPs (n = 1037). The stool sample analyses revealed the presence of multiple pathogens (rotavirus A, norovirus genogroups I and II and C. coli) as single or co-infections. Except for C. coli, these pathogens have been implicated in numerous waterborne outbreaks, suggesting an influx of sewage into the water supply system [10,12–18]. Rotaviruses are known to be responsible for large epidemics during winter seasons, affecting particularly children aged < 3 years. These epidemics are caused by the faecal route of contamination (personto-person transmission). The high proportion of stools positive for rotavirus (71%) is uncommon in waterborne outbreaks and probably reflects massive contamination [15,16,19]. Different molecular sequences of rotavirus A genotype G1 found in stools and water indicated a human faecal source of infection. In the present study, contamination of water by sewage could be linked with the large amount of sewage from the water purification plant sprayed on to land close to the river and the water catchments. In this outbreak, C. coli, which was identified in 32% of stools, was the second most frequent pathogen. To our knowledge, human C. coli contamination has not been described previously as a causative agent in waterborne outbreaks, unlike C. jejuni [11,20–25], although C. coli has been isolated from water [26]. The main symptoms observed (diarrhoea, abdominal pain, nausea) and the mean duration of the disease (4.5 days) were compatible with a Campylobacter infection. No bloody diarrhoea was reported, although bloody diarrhoea is not constant in Campylobacter infections [27]. The high proportion of individuals who had to take sick leave, and the fact that six individuals were hospitalised, suggested a relative severity of symptoms, which is much more compatible with Campylobacter infection than with virus infections. Among the hospitalised patients, 60% were infected with C. coli. Among patients whose stools were analysed and whose age was known, cases infected only with rotavirus were younger (mean age 49 years; range, 5 months to 95 years) than cases infected only with Campylobacter (mean age 66 years; range 23–88 years) or co-infected with rotavirus plus Campylobacter (mean age 69 years; range 2–91 years). Moreover, the proportion (52%) of medical visits was unusually high compared with the proportion of visits observed during other waterborne outbreaks [13,16,28]. Campylobacter needs specific atmospheric conditions for growth and survival and its detection can be very difficult [29,30]. Because of a rapid physiological transformation to a viable, though non-cultivable, state, it has been suggested that conventional culture methods often fail [31]. Nevertheless, in the present study, the failure to detect C. coli in water, and the detection of noroviruses in stools, could be caused by the small (10 mL) water samples examined. Ha¨nninen et al. [30] proposed that the volume used for detection of suspected pathogens in drinking water should be at least 8–10 L. Groundwater is considered widely to be microbiologically clean, and is usually used for drinking without treatment. However, groundwater can be contaminated by the environment [26,32]. Stanley et al. [33] showed that Campylobacter can occur in groundwater, with hydrological evidence suggesting that the source of contamination in their study was a dairy farm situated within the hydrological catchment area of the groundwater, which was confirmed when identical strains of C. jejuni were isolated from groundwater and the dairy herd. In the present study, Campylobacter was not identified from groundwater, and dairy herds and other domestic animals in farms around the source were not investigated. However, results of bacterial, chemical and physical water analyses suggested strongly that a river contaminated by agricultural run-off from surrounding sheep and cattle grazing meadows was contaminating the source. In addition, the immediate area of the source was insufficiently protected and vulnerable to the environment. Wild birds, rodents, pigs and cows are all hosts for C. coli and could contaminate surface water easily [26,27,34,35]. Campylobacters are sensitive to many environmental factors and to chlorine [33,36,37], but a failure of the chlorination system could have resulted in the presence of a high concentration of C. coli. Waterborne outbreaks are often caused by several factors that act together [18,20,23,38]. In some outbreaks, heavy rainfall has been responsible for increasing agricultural run-off into rivers, although this did not occur before the outbreak in Gourdon. This underlies the need for increased frequency of routine water quality monitoring, chlorination of the water supply according to environmental conditions, and implementation of preventive measures. This outbreak raised the question of the delay in raising the alert. An outbreak can be recognised by: (i) the poor quality of water detected during regular controls; (ii) complaints about water quality; (iii) an incident in the supply and ℠or sewage water system during maintenance work; and (iv) an increased number of cases of gastroenteritis. In most instances, waterborne outbreaks are suspected when there are already a large number of declared cases [3,12,18,28,38,39]. In the present outbreak, the initial warning was given by a GP who notified officials 5–10 days after the assumed date of onset. This highlights the role of GPs in the detection of an unusual event, particularly during the summer, when an increase in acute gastroenteritis is less expected than during winter [40,41]. In addition, a routine water quality monitoring system able to detect contamination of drinking water before the occurrence of illness and complaints about the water quality should be used as an early warning source. The retrospective cohort study of Gourdon residents allowed the scale and severity of the outbreak to be quantified precisely and the hypothesis that tap water was the vehicle of the outbreak to be tested. When provided with a cohort sample representative of the study population, the number of people affected can be estimated. However, as the exposure of the population was high, a relatively large sample size was needed to test the association between the disease and the exposure. Because of misclassification of exposure (unrecognised consumption of tap water), the attributable risk is probably underestimated. In the absence of obvious alternative explanations, water contamination is the most likely common cause of an explosive outbreak of AGE occurring in a large population. Tillett et al. [42] proposed categorising levels of evidence for waterborne outbreaks. A descriptive study suggesting a water-related outbreak, together with the identification of the same pathogen in stools and water, should be sufficient to conclude that an outbreak is associated strongly with water consumption. Faecal indicators of water quality are currently insufficiently sensitive for some pathogens, notably viruses and parasites (e.g., Cryptosporidium). Therefore, in the absence of evidence from an analytical study to demonstrate an association, the link between illness and water can only be classified as probable or possible. However, total coliform counts are good indicators of faecal contamination with other pathogens (e.g., rotaviruses, E. coli, Campylobacter) [43]. In conclusion, the waterborne outbreak described in the present study was most probably caused by faecal contamination of source A by groundwater, resulting from deficiencies in the maintenance of the water distribution system. Corrective measures were implemented based on the results of the investigation. To our knowledge, this is the first documented waterborne outbreak involving human C. coli infection. In-depth environmental investigations to identify the origin of water contamination are important in order to better understand microbial transmission from the environment and to implement appropriate controls and preventive measures. In addition, surveys on vehicles and vectors that transmit Campylobacter between hosts are crucial in order to better understand the epidemiology of these organisms.

When did this event start?

280

A large multi-pathogen waterborne community outbreak linked to faecal contamination of a groundwater system, France, 2000

A large waterborne outbreak of infection that occurred during August 2000 in a local community in France was investigated initially via a rapid survey of visits to local physicians. A retrospective cohort study was then conducted on a random cluster sample of residents. Of 709 residents interviewed, 202 (28.5%) were definite cases (at least three liquid stools ℠day or vomiting) and 62 (8.7%) were probable cases (less than three liquid stools ℠day or abdominal pain). Those who had drunk tap water had a threefold increased risk for illness (95% CI 2.4–4.0). The risk increased with the amount of water consumed (chi-square trend: p < 0.0001). Bacteriological analyses of stools were performed for 35 patients and virological analyses for 24 patients. Campylobacter coli, group A rotavirus and norovirus were detected in 31.5%, 71.0% and 21% of samples, respectively. An extensive environmental investigation concluded that a groundwater source to this community had probably been contaminated by agricultural run-off, and a failure in the chlorination system was identified. This is the first documented waterborne outbreak of infection involving human C. coli infections. A better understanding of the factors influencing campylobacter transmission between hosts is required. Keywords Campylobacter coli, epidemiology, gastroenteritis, norovirus, rotavirus, waterborne outbreak INTRODUCTION Contaminated drinking water causes extensive outbreaks of illness because of the large number of people served by water supply facilities. Detection of such outbreaks requires the identification of an increase in illness (usually the rate of gastrointestinal disease) in the exposed population and confirmation that water was the route of transmission [1]. Many countries have routine reporting systems for detecting foodborne and waterborne outbreaks of infection. However, in routine practice it is not easy to detect an increase in clinical cases and to link this increase to waterborne transmission; thus waterborne outbreaks are often unrecognised and underestimated [2]. Microbiological examination of water is more complex than examination of stools, and there is often a failure to detect pathogens in water. Norovirus, Cryptosporidium parvum, Giardia intestinalis and Campylobacter jejuni are the pathogens identified most frequently [3]. The present study describes a large waterborne community outbreak of infection with multiple pathogens, including Campylobacter coli. METHODS AND MATERIALS Background On 23 August 2000, a general practitioner (GP) informed the local district health department that 16 cases of acute gastroenteritis (AGE) had occurred among residents of a holiday camp in the Gourdon community since 21 August. Gourdon, which is located in the south-west of France, has a population of 4888 inhabitants, which increases by 50% in the July– August tourist season. Investigations at the holiday camp concluded that foodborne transmission was very unlikely, but found that tap water was highly contaminated with faecal coliforms. Preliminary information collected from physicians in Gourdon indicated an increase of the number of consultations for AGE. Gourdon tap water is supplied by two underground water facilities: facility A serving 3800 water consumers in Gourdon; and facility B serving 1200 water consumers in Gourdon and the surrounding communities. Water treatment in facility A consists of chlorination, which involves two automatic pumps working in rotation. The water then passes through two semi in-ground reservoir tanks (1100 m3 and 200 m3, respectively) before distribution. Facility B water is mixed with a fraction of facility A water in another 500 m3 semi in-ground reservoir tank before being distributed. This mixing with the treated water from facility A is intended to disinfect the water from facility B. Since the increase in AGE was noted only in Gourdon, contaminated tap water from facility A was suspected to be the origin of the outbreak. Therefore, the population of Gourdon was informed on 25 August to avoid the consumption of tap water or to boil it for 5 min. Survey of gastroenteritis-related medical consultations To confirm the outbreak, the 11 GPs in Gourdon and the local hospital emergency department were requested to provide the number of total daily visits and the number of cases of gastroenteritis cases occurring between 1 August and 11 September 2000. Retrospective cohort study A retrospective cohort study was performed on a sample of Gourdon households between 5 and 15 September 2000. Domestic households were selected randomly from the telephone directory. A definite case was defined as a Gourdon resident who had at least three liquid stools ℠day or vomiting between 1 August and 3 September, and who had been present in Gourdon for ‡ 1 day between 1 and 31 August; if less than three liquid stools ℠day or abdominal pain had occurred, the case was classified as probable. Exposure was defined as the consumption of tap water before 25 August 2000. A standardised telephone questionnaire was completed to obtain demographic data (gender, date of birth), food consumption (commercially prepared dishes, sea-food, raw milk, pastry), tap water consumption (before and after 25 August, daily number of glasses of water, consumption of boiled tap water), symptoms of gastroenteritis (date of onset, number of liquid stools ℠day, vomiting, nausea, abdominal pain, fever), medical care (in- or out-patient) and consequences on daily activities (sick leave, bed confinement) for each member of the selected household. Under the hypothesis that 60% of the population consumed tap water, an estimated 550 people were required for inclusion in the study to identify a relative risk (RR) of ‡ 3, with a 5% a risk and a power of 80%. Based on an average of two individuals per household (INSERM 1999 population census), 300 households were required for inclusion in the survey. The association between AGE and the consumption of tap water was assessed by calculating the risk ratio (RR) and 95% CI, taking into account a household cluster design effect. A dose-effect relationship was evaluated by analysing the RR trend for different categories of daily number of glasses of tap water consumed (chi-square trend). Proportions were compared using the corrected chi-square test, and means were compared using the Student test or Mann and Witney test. Data were analysed with Epi-Info 6.04c (CDC, Atlanta, GA, USA). The gender and age distribution of the cohort population were compared with that of the general population of Gourdon (INSERM 1999 population census) to assess whether it was representative. Under the hypothesis of representation, the total number of cases of gastroenteritis was estimated by applying the attack rate to the total population of Gourdon, including the tourist population (data provided by the Gourdon tourism office). The most likely period for contamination of the water distribution system was based on the epidemic curve and the minimal incubation period of the pathogens involved. Microbiological analyses of stools Stool samples were examined for the presence of Salmonella spp., Shigella spp., Staphylococcus aureus, Campylobacter spp., Yersinia enterocolitica, Escherichia coli, enterococci, rotavirus groups A and C, astrovirus, calicivirus, adenovirus types 40 and 41, enterovirus, hepatitis A virus and Cryptosporidum. Viruses were detected using an enzyme immunoassay for group A rotavirus, astrovirus and adenovirus 40 ℠41, and by RT-PCR for calicivirus, group C rotavirus, enterovirus and hepatitis A virus [4–6]. Campylobacter isolates were sent to the National Reference Centre for Campylobacter for identification and fingerprinting. Species identification was by means of phenotypic tests (API Campy; bioMe´rieux, Lyon, France), with use of PCR to differentiate Campylobacter jejuni from Camplylobacter coli [7]. Random amplified polymorphic DNA fingerprinting was performed with primer 3881, 5¢-AACGCGCAAC. Environmental investigation An environmental and microbiological investigation of the water distribution system was performed. The inspection concerned the water catchments, the water distribution and sewage system, agricultural practices and private households around the sources. Bacteriological (Escherichia coli, enterococci, faecal streptococci, thermotolerant coliforms, sulphite-reducing clostridia) and chemical (chlorine) analyses were performed on tap water and on the water sampled from the two groundwater sources. After 23 August, the water chlorine concentration and the presence of coliform bacteria were monitored daily at several sites of the water system before and after the resort treatment. Examinations for viruses (group A rotavirus, astrovirus, norovirus, enterovirus, hepatitis A virus) and Campylobacter spp. were undertaken on two 10-mL tap water samples collected inhouse on 23 August, and on two 10-mL samples from groundwater sources A and B on 25 August. Viruses were detected by RT-PCR, followed by hybridisation with a specific primer to improve sensitivity, while C. jejuni and C. coli were detected by PCR with specific primers. Group A rotavirus isolates detected in stools and in water were sequenced using Beg 9 and End 9 primers [8] and compared using the Infobiogene GCG software (http://www.infobiogen.fr). Bacteriological (revived 36C, revived 22C, thermotolerant coliforms, streptococci and faecal enterococci), physical (temperature, conductivity) and chemical (oxygen consumption, ammonium, nitrates, phosphorus) parameters from source A and the river were analysed. Trends were compared for an 8-week period between 21 December 2000 and 7 February 2001. RESULTS Epidemiology Between 1 August and 11 September 2000, 7104 Gourdon residents consulted a GP or the local hospital emergency department. Between 24 and 28 August, gastroenteritis accounted for 44% (479 ℠1097) of all medical visits, compared with 6% (126 ℠2185) during the first 15 days of August (Fig. 1). Among the 498 households contacted for the household survey, 198 were excluded (103 did not respond, 49 did not fulfil the inclusion criteria, 43 refused to participate, and three for unknown reasons). The questionnaire was completed for 300 households (709 individuals). Data were missing concerning gender for one individual and age for 11 individuals. Overall, 331 (46.7%; 95% CI 44.5–49.0) of respondents were male and 377 (53.2%; 95% CI 50.9–55.5) were female. The median age was 44 years (range < 1–94 years). The age and gender distribution of the sample was similar to that of the general population. During the outbreak period, 264 of the 709 respondents (attack rate (AR) 37.2%) had been ill, with 202 definite cases (AR 28.5%; 95% CI 24.6– 32.4) and 62 probable cases (AR 8.7%; 95% CI 6.5– 11.0). The AR was higher in females (31.8%) than in males (24.8%), and was highest in children aged < 6 years (42.1%, Table 1). The epidemic curve (Fig. 2) suggests that the outbreak started between 14 and 19 August. The shape of the epidemic curve in the retrospective cohort study was very similar to that based on the number of visits to GPs for gastroenteritis. Patients complained mainly of diarrhoea, abdominal pain and nausea (Table 2). No cases of bloody diarrhoea were notified. The mean duration of symptoms was 4.5 days (median 3, range 1–30 days). Among definite cases, 52% had visited a GP, 42% had taken sick leave for ‡ 1 day, 32% had required bed rest, and 2.9% (six patients) had been hospitalised. In total, 336 (47.4%) respondents had drunk tap water before 25 August, whereas 646 (91.1%) did not drink tap water after 25 August. The risk for illness, taking into account the sample cluster design effect of 1.4, was three-fold greater (95% CI 2.2–4.1) for those who had drunk tap water before 25 August, and increased with the amount of water consumed (chi-square trend: p < 0.0001; Table 3). The risk of gastroenteritis associated with drinking tap water did not change substantially after adjustment for age (RRMH 3.3, 95% CI 2.5–4.4), and there was no modification effect according to age. Food consumption was not associated with illness. Applying the AR of 37% to the total population of Gourdon exposed to the water supply (4888 residents + 2200 tourists), the number of individuals affected by this outbreak was estimated to be 2600 (95% CI 2400–2900), assuming an exposure to tap water and a risk for gastroenteritis among the tourists similar to that among the inhabitants. Microbiological stool analyses Among the 35 stool samples analysed for bacteria, 11 (31.5%) were positive for C. coli. Twenty-four of the 35 samples were also tested for viruses: rotavirus A was detected in 17 (71.0%) and norovirus in five (21.0%) samples, while nine (37.5%) samples were co-infected (four C. coli + rotavirus; one C. coli + norovirus; three rotavirus + norovirus; and one C. coli + rotavirus + norovirus). Of the 11 C. coli isolates, six were characterised as biotype 2, and were identical following RAPD analysis, while the remaining five were of biotype 1. Rotavirus typing by RTPCR identified a single genotype P-(8), G1. Molecular characterisation of the norovirus isolates identified two different patterns, with three isolates being identical and close to the Saragota strain (genogroup I) and two close to the new GGIIb variant (genogroup II). Environmental investigation All of the water samples from source B were negative, regardless of the site of the distribution system tested. Analyses of the water originating from source A, sampled after leaving the water-works on 23 and 25 August, revealed high concentrations of E. coli (2263 / 100 mL), enterococci (415 / 100 mL), total coliforms (90 / 100 mL), faecal streptococci (136 / 100 mL) and sulphite-reducing clostridia (10 / 100 mL), and an absence of chlorination. Rotavirus A genotype G1 was detected in the crude water sampled from source A. The analyses for enterovirus, astrovirus, norovirus, hepatitis A virus and Campylobacter were negative. Source A was situated in a valley containing a river, below a hamlet of 15 dwellings, including farms breeding cattle, pork and sheep. The source was not supposed to be supplied by the river. The river resulted from the convergence of about ten different streams, with one originating close to a slaughterhouse, and was exposed to agricultural run-off and manure effluent; sheep and cattle grazed in meadows bordering the river, and a large amount of sewage from water treatment facilities was sprayed on to land close to the river. Cowsheds were close to the water catchment area. The immediate protected area around source A was too small and littered with building rubbish. Air vents without protection in the door made the source accessible to animals and insects. Bacteriological analyses of the water samples from source A and the river revealed that the contamination was of both human and animal origin, with a higher concentration in the river. The levels and trends of the chemical parameters of source A were similar to those of the river, strongly suggesting that the river was connected to source A (Fig. 3). A retrospective review of routine data collected between 1991 and 2000 showed irregularities in the maintenance of the water distribution system and the water treatment plant, particularly during the month of August. In the hamlet, the sewage system of the 15 households did not comply with sanitary rules. All but three ejected waste either into a cesspool or superficially without treatment. Three regions of gene 9 of the rotavirus detected in stools and water were compared. Sequencing showed that the strains were closely related but not identical. Assuming a likely onset of the outbreak between 14 and 19 August, and taking into account a minimum incubation period of 2 days for rotavirus A, norovirus and C. coli, the contamination of the water distribution system probably occurred between 12 and 17 August. Contamination probably persisted until 28 August, as intensive chlorination, beginning on 25 August, was effective only from 28 August onwards (Fig. 2). DISCUSSION The present descriptive and retrospective cohort studies confirmed that the outbreak of infection in the community of Gourdon was related to the consumption of tap water and affected 2600 individuals. The outbreak involved multiple pathogens, with group A rotavirus and C. coli being the most frequent, followed by norovirus. Results of environmental investigations suggested that water contamination was related to the contamination of the source, combined with a failure of the water chlorination system. The association with tap water consumption, the dose-effect relationship, and the fact that multiple pathogens were involved, were all strongly in favour of tap water as the origin of infection. Moreover, the number of cases decreased dramatically after the water restriction advice and the implementation of control measures on the water supply. No food products explored in the study were associated with illness. An AR of 14.5% and 31% was observed among adults and children aged < 6 years, respectively, who reported that they had not consumed tap water. This relatively high rate could be explained by several phenomena. Many individuals may have consumed tap water without being aware of it, e.g., in drinks or ice cubes consumed outside of the home, via raw fruits and vegetables, dishes prepared with unboiled water, or while cleaning their teeth. It can be hypothesised that people who became ill without being aware of having consumed tap water were exposed to small amounts of water and had been subjected to a small infective dose. In addition, children, more than adults, may have been contaminated through person-to-person transmission (e.g., family contact), which is the most common route of transmission for rotavirus and calicivirus. Waterborne contamination can affect large numbers of individuals [3,9–11]. In the present study, assuming a 50% increase in the population during the tourist season, the number of affected people was estimated to be between 2400 and 2900. The cohort study showed that almost half (46.5%) of the cases consulted a GP. Applying this proportion to the estimated number of cases yielded an estimated 1116–1349 ill individuals who consulted a GP for gastroenteritis, similar to the number of consultations for gastroenteritis actually reported by the GPs (n = 1037). The stool sample analyses revealed the presence of multiple pathogens (rotavirus A, norovirus genogroups I and II and C. coli) as single or co-infections. Except for C. coli, these pathogens have been implicated in numerous waterborne outbreaks, suggesting an influx of sewage into the water supply system [10,12–18]. Rotaviruses are known to be responsible for large epidemics during winter seasons, affecting particularly children aged < 3 years. These epidemics are caused by the faecal route of contamination (personto-person transmission). The high proportion of stools positive for rotavirus (71%) is uncommon in waterborne outbreaks and probably reflects massive contamination [15,16,19]. Different molecular sequences of rotavirus A genotype G1 found in stools and water indicated a human faecal source of infection. In the present study, contamination of water by sewage could be linked with the large amount of sewage from the water purification plant sprayed on to land close to the river and the water catchments. In this outbreak, C. coli, which was identified in 32% of stools, was the second most frequent pathogen. To our knowledge, human C. coli contamination has not been described previously as a causative agent in waterborne outbreaks, unlike C. jejuni [11,20–25], although C. coli has been isolated from water [26]. The main symptoms observed (diarrhoea, abdominal pain, nausea) and the mean duration of the disease (4.5 days) were compatible with a Campylobacter infection. No bloody diarrhoea was reported, although bloody diarrhoea is not constant in Campylobacter infections [27]. The high proportion of individuals who had to take sick leave, and the fact that six individuals were hospitalised, suggested a relative severity of symptoms, which is much more compatible with Campylobacter infection than with virus infections. Among the hospitalised patients, 60% were infected with C. coli. Among patients whose stools were analysed and whose age was known, cases infected only with rotavirus were younger (mean age 49 years; range, 5 months to 95 years) than cases infected only with Campylobacter (mean age 66 years; range 23–88 years) or co-infected with rotavirus plus Campylobacter (mean age 69 years; range 2–91 years). Moreover, the proportion (52%) of medical visits was unusually high compared with the proportion of visits observed during other waterborne outbreaks [13,16,28]. Campylobacter needs specific atmospheric conditions for growth and survival and its detection can be very difficult [29,30]. Because of a rapid physiological transformation to a viable, though non-cultivable, state, it has been suggested that conventional culture methods often fail [31]. Nevertheless, in the present study, the failure to detect C. coli in water, and the detection of noroviruses in stools, could be caused by the small (10 mL) water samples examined. Ha¨nninen et al. [30] proposed that the volume used for detection of suspected pathogens in drinking water should be at least 8–10 L. Groundwater is considered widely to be microbiologically clean, and is usually used for drinking without treatment. However, groundwater can be contaminated by the environment [26,32]. Stanley et al. [33] showed that Campylobacter can occur in groundwater, with hydrological evidence suggesting that the source of contamination in their study was a dairy farm situated within the hydrological catchment area of the groundwater, which was confirmed when identical strains of C. jejuni were isolated from groundwater and the dairy herd. In the present study, Campylobacter was not identified from groundwater, and dairy herds and other domestic animals in farms around the source were not investigated. However, results of bacterial, chemical and physical water analyses suggested strongly that a river contaminated by agricultural run-off from surrounding sheep and cattle grazing meadows was contaminating the source. In addition, the immediate area of the source was insufficiently protected and vulnerable to the environment. Wild birds, rodents, pigs and cows are all hosts for C. coli and could contaminate surface water easily [26,27,34,35]. Campylobacters are sensitive to many environmental factors and to chlorine [33,36,37], but a failure of the chlorination system could have resulted in the presence of a high concentration of C. coli. Waterborne outbreaks are often caused by several factors that act together [18,20,23,38]. In some outbreaks, heavy rainfall has been responsible for increasing agricultural run-off into rivers, although this did not occur before the outbreak in Gourdon. This underlies the need for increased frequency of routine water quality monitoring, chlorination of the water supply according to environmental conditions, and implementation of preventive measures. This outbreak raised the question of the delay in raising the alert. An outbreak can be recognised by: (i) the poor quality of water detected during regular controls; (ii) complaints about water quality; (iii) an incident in the supply and ℠or sewage water system during maintenance work; and (iv) an increased number of cases of gastroenteritis. In most instances, waterborne outbreaks are suspected when there are already a large number of declared cases [3,12,18,28,38,39]. In the present outbreak, the initial warning was given by a GP who notified officials 5–10 days after the assumed date of onset. This highlights the role of GPs in the detection of an unusual event, particularly during the summer, when an increase in acute gastroenteritis is less expected than during winter [40,41]. In addition, a routine water quality monitoring system able to detect contamination of drinking water before the occurrence of illness and complaints about the water quality should be used as an early warning source. The retrospective cohort study of Gourdon residents allowed the scale and severity of the outbreak to be quantified precisely and the hypothesis that tap water was the vehicle of the outbreak to be tested. When provided with a cohort sample representative of the study population, the number of people affected can be estimated. However, as the exposure of the population was high, a relatively large sample size was needed to test the association between the disease and the exposure. Because of misclassification of exposure (unrecognised consumption of tap water), the attributable risk is probably underestimated. In the absence of obvious alternative explanations, water contamination is the most likely common cause of an explosive outbreak of AGE occurring in a large population. Tillett et al. [42] proposed categorising levels of evidence for waterborne outbreaks. A descriptive study suggesting a water-related outbreak, together with the identification of the same pathogen in stools and water, should be sufficient to conclude that an outbreak is associated strongly with water consumption. Faecal indicators of water quality are currently insufficiently sensitive for some pathogens, notably viruses and parasites (e.g., Cryptosporidium). Therefore, in the absence of evidence from an analytical study to demonstrate an association, the link between illness and water can only be classified as probable or possible. However, total coliform counts are good indicators of faecal contamination with other pathogens (e.g., rotaviruses, E. coli, Campylobacter) [43]. In conclusion, the waterborne outbreak described in the present study was most probably caused by faecal contamination of source A by groundwater, resulting from deficiencies in the maintenance of the water distribution system. Corrective measures were implemented based on the results of the investigation. To our knowledge, this is the first documented waterborne outbreak involving human C. coli infection. In-depth environmental investigations to identify the origin of water contamination are important in order to better understand microbial transmission from the environment and to implement appropriate controls and preventive measures. In addition, surveys on vehicles and vectors that transmit Campylobacter between hosts are crucial in order to better understand the epidemiology of these organisms.

When did this event start?

281

A large multi-pathogen waterborne community outbreak linked to faecal contamination of a groundwater system, France, 2000

A large waterborne outbreak of infection that occurred during August 2000 in a local community in France was investigated initially via a rapid survey of visits to local physicians. A retrospective cohort study was then conducted on a random cluster sample of residents. Of 709 residents interviewed, 202 (28.5%) were definite cases (at least three liquid stools ℠day or vomiting) and 62 (8.7%) were probable cases (less than three liquid stools ℠day or abdominal pain). Those who had drunk tap water had a threefold increased risk for illness (95% CI 2.4–4.0). The risk increased with the amount of water consumed (chi-square trend: p < 0.0001). Bacteriological analyses of stools were performed for 35 patients and virological analyses for 24 patients. Campylobacter coli, group A rotavirus and norovirus were detected in 31.5%, 71.0% and 21% of samples, respectively. An extensive environmental investigation concluded that a groundwater source to this community had probably been contaminated by agricultural run-off, and a failure in the chlorination system was identified. This is the first documented waterborne outbreak of infection involving human C. coli infections. A better understanding of the factors influencing campylobacter transmission between hosts is required. Keywords Campylobacter coli, epidemiology, gastroenteritis, norovirus, rotavirus, waterborne outbreak INTRODUCTION Contaminated drinking water causes extensive outbreaks of illness because of the large number of people served by water supply facilities. Detection of such outbreaks requires the identification of an increase in illness (usually the rate of gastrointestinal disease) in the exposed population and confirmation that water was the route of transmission [1]. Many countries have routine reporting systems for detecting foodborne and waterborne outbreaks of infection. However, in routine practice it is not easy to detect an increase in clinical cases and to link this increase to waterborne transmission; thus waterborne outbreaks are often unrecognised and underestimated [2]. Microbiological examination of water is more complex than examination of stools, and there is often a failure to detect pathogens in water. Norovirus, Cryptosporidium parvum, Giardia intestinalis and Campylobacter jejuni are the pathogens identified most frequently [3]. The present study describes a large waterborne community outbreak of infection with multiple pathogens, including Campylobacter coli. METHODS AND MATERIALS Background On 23 August 2000, a general practitioner (GP) informed the local district health department that 16 cases of acute gastroenteritis (AGE) had occurred among residents of a holiday camp in the Gourdon community since 21 August. Gourdon, which is located in the south-west of France, has a population of 4888 inhabitants, which increases by 50% in the July– August tourist season. Investigations at the holiday camp concluded that foodborne transmission was very unlikely, but found that tap water was highly contaminated with faecal coliforms. Preliminary information collected from physicians in Gourdon indicated an increase of the number of consultations for AGE. Gourdon tap water is supplied by two underground water facilities: facility A serving 3800 water consumers in Gourdon; and facility B serving 1200 water consumers in Gourdon and the surrounding communities. Water treatment in facility A consists of chlorination, which involves two automatic pumps working in rotation. The water then passes through two semi in-ground reservoir tanks (1100 m3 and 200 m3, respectively) before distribution. Facility B water is mixed with a fraction of facility A water in another 500 m3 semi in-ground reservoir tank before being distributed. This mixing with the treated water from facility A is intended to disinfect the water from facility B. Since the increase in AGE was noted only in Gourdon, contaminated tap water from facility A was suspected to be the origin of the outbreak. Therefore, the population of Gourdon was informed on 25 August to avoid the consumption of tap water or to boil it for 5 min. Survey of gastroenteritis-related medical consultations To confirm the outbreak, the 11 GPs in Gourdon and the local hospital emergency department were requested to provide the number of total daily visits and the number of cases of gastroenteritis cases occurring between 1 August and 11 September 2000. Retrospective cohort study A retrospective cohort study was performed on a sample of Gourdon households between 5 and 15 September 2000. Domestic households were selected randomly from the telephone directory. A definite case was defined as a Gourdon resident who had at least three liquid stools ℠day or vomiting between 1 August and 3 September, and who had been present in Gourdon for ‡ 1 day between 1 and 31 August; if less than three liquid stools ℠day or abdominal pain had occurred, the case was classified as probable. Exposure was defined as the consumption of tap water before 25 August 2000. A standardised telephone questionnaire was completed to obtain demographic data (gender, date of birth), food consumption (commercially prepared dishes, sea-food, raw milk, pastry), tap water consumption (before and after 25 August, daily number of glasses of water, consumption of boiled tap water), symptoms of gastroenteritis (date of onset, number of liquid stools ℠day, vomiting, nausea, abdominal pain, fever), medical care (in- or out-patient) and consequences on daily activities (sick leave, bed confinement) for each member of the selected household. Under the hypothesis that 60% of the population consumed tap water, an estimated 550 people were required for inclusion in the study to identify a relative risk (RR) of ‡ 3, with a 5% a risk and a power of 80%. Based on an average of two individuals per household (INSERM 1999 population census), 300 households were required for inclusion in the survey. The association between AGE and the consumption of tap water was assessed by calculating the risk ratio (RR) and 95% CI, taking into account a household cluster design effect. A dose-effect relationship was evaluated by analysing the RR trend for different categories of daily number of glasses of tap water consumed (chi-square trend). Proportions were compared using the corrected chi-square test, and means were compared using the Student test or Mann and Witney test. Data were analysed with Epi-Info 6.04c (CDC, Atlanta, GA, USA). The gender and age distribution of the cohort population were compared with that of the general population of Gourdon (INSERM 1999 population census) to assess whether it was representative. Under the hypothesis of representation, the total number of cases of gastroenteritis was estimated by applying the attack rate to the total population of Gourdon, including the tourist population (data provided by the Gourdon tourism office). The most likely period for contamination of the water distribution system was based on the epidemic curve and the minimal incubation period of the pathogens involved. Microbiological analyses of stools Stool samples were examined for the presence of Salmonella spp., Shigella spp., Staphylococcus aureus, Campylobacter spp., Yersinia enterocolitica, Escherichia coli, enterococci, rotavirus groups A and C, astrovirus, calicivirus, adenovirus types 40 and 41, enterovirus, hepatitis A virus and Cryptosporidum. Viruses were detected using an enzyme immunoassay for group A rotavirus, astrovirus and adenovirus 40 ℠41, and by RT-PCR for calicivirus, group C rotavirus, enterovirus and hepatitis A virus [4–6]. Campylobacter isolates were sent to the National Reference Centre for Campylobacter for identification and fingerprinting. Species identification was by means of phenotypic tests (API Campy; bioMe´rieux, Lyon, France), with use of PCR to differentiate Campylobacter jejuni from Camplylobacter coli [7]. Random amplified polymorphic DNA fingerprinting was performed with primer 3881, 5¢-AACGCGCAAC. Environmental investigation An environmental and microbiological investigation of the water distribution system was performed. The inspection concerned the water catchments, the water distribution and sewage system, agricultural practices and private households around the sources. Bacteriological (Escherichia coli, enterococci, faecal streptococci, thermotolerant coliforms, sulphite-reducing clostridia) and chemical (chlorine) analyses were performed on tap water and on the water sampled from the two groundwater sources. After 23 August, the water chlorine concentration and the presence of coliform bacteria were monitored daily at several sites of the water system before and after the resort treatment. Examinations for viruses (group A rotavirus, astrovirus, norovirus, enterovirus, hepatitis A virus) and Campylobacter spp. were undertaken on two 10-mL tap water samples collected inhouse on 23 August, and on two 10-mL samples from groundwater sources A and B on 25 August. Viruses were detected by RT-PCR, followed by hybridisation with a specific primer to improve sensitivity, while C. jejuni and C. coli were detected by PCR with specific primers. Group A rotavirus isolates detected in stools and in water were sequenced using Beg 9 and End 9 primers [8] and compared using the Infobiogene GCG software (http://www.infobiogen.fr). Bacteriological (revived 36C, revived 22C, thermotolerant coliforms, streptococci and faecal enterococci), physical (temperature, conductivity) and chemical (oxygen consumption, ammonium, nitrates, phosphorus) parameters from source A and the river were analysed. Trends were compared for an 8-week period between 21 December 2000 and 7 February 2001. RESULTS Epidemiology Between 1 August and 11 September 2000, 7104 Gourdon residents consulted a GP or the local hospital emergency department. Between 24 and 28 August, gastroenteritis accounted for 44% (479 ℠1097) of all medical visits, compared with 6% (126 ℠2185) during the first 15 days of August (Fig. 1). Among the 498 households contacted for the household survey, 198 were excluded (103 did not respond, 49 did not fulfil the inclusion criteria, 43 refused to participate, and three for unknown reasons). The questionnaire was completed for 300 households (709 individuals). Data were missing concerning gender for one individual and age for 11 individuals. Overall, 331 (46.7%; 95% CI 44.5–49.0) of respondents were male and 377 (53.2%; 95% CI 50.9–55.5) were female. The median age was 44 years (range < 1–94 years). The age and gender distribution of the sample was similar to that of the general population. During the outbreak period, 264 of the 709 respondents (attack rate (AR) 37.2%) had been ill, with 202 definite cases (AR 28.5%; 95% CI 24.6– 32.4) and 62 probable cases (AR 8.7%; 95% CI 6.5– 11.0). The AR was higher in females (31.8%) than in males (24.8%), and was highest in children aged < 6 years (42.1%, Table 1). The epidemic curve (Fig. 2) suggests that the outbreak started between 14 and 19 August. The shape of the epidemic curve in the retrospective cohort study was very similar to that based on the number of visits to GPs for gastroenteritis. Patients complained mainly of diarrhoea, abdominal pain and nausea (Table 2). No cases of bloody diarrhoea were notified. The mean duration of symptoms was 4.5 days (median 3, range 1–30 days). Among definite cases, 52% had visited a GP, 42% had taken sick leave for ‡ 1 day, 32% had required bed rest, and 2.9% (six patients) had been hospitalised. In total, 336 (47.4%) respondents had drunk tap water before 25 August, whereas 646 (91.1%) did not drink tap water after 25 August. The risk for illness, taking into account the sample cluster design effect of 1.4, was three-fold greater (95% CI 2.2–4.1) for those who had drunk tap water before 25 August, and increased with the amount of water consumed (chi-square trend: p < 0.0001; Table 3). The risk of gastroenteritis associated with drinking tap water did not change substantially after adjustment for age (RRMH 3.3, 95% CI 2.5–4.4), and there was no modification effect according to age. Food consumption was not associated with illness. Applying the AR of 37% to the total population of Gourdon exposed to the water supply (4888 residents + 2200 tourists), the number of individuals affected by this outbreak was estimated to be 2600 (95% CI 2400–2900), assuming an exposure to tap water and a risk for gastroenteritis among the tourists similar to that among the inhabitants. Microbiological stool analyses Among the 35 stool samples analysed for bacteria, 11 (31.5%) were positive for C. coli. Twenty-four of the 35 samples were also tested for viruses: rotavirus A was detected in 17 (71.0%) and norovirus in five (21.0%) samples, while nine (37.5%) samples were co-infected (four C. coli + rotavirus; one C. coli + norovirus; three rotavirus + norovirus; and one C. coli + rotavirus + norovirus). Of the 11 C. coli isolates, six were characterised as biotype 2, and were identical following RAPD analysis, while the remaining five were of biotype 1. Rotavirus typing by RTPCR identified a single genotype P-(8), G1. Molecular characterisation of the norovirus isolates identified two different patterns, with three isolates being identical and close to the Saragota strain (genogroup I) and two close to the new GGIIb variant (genogroup II). Environmental investigation All of the water samples from source B were negative, regardless of the site of the distribution system tested. Analyses of the water originating from source A, sampled after leaving the water-works on 23 and 25 August, revealed high concentrations of E. coli (2263 / 100 mL), enterococci (415 / 100 mL), total coliforms (90 / 100 mL), faecal streptococci (136 / 100 mL) and sulphite-reducing clostridia (10 / 100 mL), and an absence of chlorination. Rotavirus A genotype G1 was detected in the crude water sampled from source A. The analyses for enterovirus, astrovirus, norovirus, hepatitis A virus and Campylobacter were negative. Source A was situated in a valley containing a river, below a hamlet of 15 dwellings, including farms breeding cattle, pork and sheep. The source was not supposed to be supplied by the river. The river resulted from the convergence of about ten different streams, with one originating close to a slaughterhouse, and was exposed to agricultural run-off and manure effluent; sheep and cattle grazed in meadows bordering the river, and a large amount of sewage from water treatment facilities was sprayed on to land close to the river. Cowsheds were close to the water catchment area. The immediate protected area around source A was too small and littered with building rubbish. Air vents without protection in the door made the source accessible to animals and insects. Bacteriological analyses of the water samples from source A and the river revealed that the contamination was of both human and animal origin, with a higher concentration in the river. The levels and trends of the chemical parameters of source A were similar to those of the river, strongly suggesting that the river was connected to source A (Fig. 3). A retrospective review of routine data collected between 1991 and 2000 showed irregularities in the maintenance of the water distribution system and the water treatment plant, particularly during the month of August. In the hamlet, the sewage system of the 15 households did not comply with sanitary rules. All but three ejected waste either into a cesspool or superficially without treatment. Three regions of gene 9 of the rotavirus detected in stools and water were compared. Sequencing showed that the strains were closely related but not identical. Assuming a likely onset of the outbreak between 14 and 19 August, and taking into account a minimum incubation period of 2 days for rotavirus A, norovirus and C. coli, the contamination of the water distribution system probably occurred between 12 and 17 August. Contamination probably persisted until 28 August, as intensive chlorination, beginning on 25 August, was effective only from 28 August onwards (Fig. 2). DISCUSSION The present descriptive and retrospective cohort studies confirmed that the outbreak of infection in the community of Gourdon was related to the consumption of tap water and affected 2600 individuals. The outbreak involved multiple pathogens, with group A rotavirus and C. coli being the most frequent, followed by norovirus. Results of environmental investigations suggested that water contamination was related to the contamination of the source, combined with a failure of the water chlorination system. The association with tap water consumption, the dose-effect relationship, and the fact that multiple pathogens were involved, were all strongly in favour of tap water as the origin of infection. Moreover, the number of cases decreased dramatically after the water restriction advice and the implementation of control measures on the water supply. No food products explored in the study were associated with illness. An AR of 14.5% and 31% was observed among adults and children aged < 6 years, respectively, who reported that they had not consumed tap water. This relatively high rate could be explained by several phenomena. Many individuals may have consumed tap water without being aware of it, e.g., in drinks or ice cubes consumed outside of the home, via raw fruits and vegetables, dishes prepared with unboiled water, or while cleaning their teeth. It can be hypothesised that people who became ill without being aware of having consumed tap water were exposed to small amounts of water and had been subjected to a small infective dose. In addition, children, more than adults, may have been contaminated through person-to-person transmission (e.g., family contact), which is the most common route of transmission for rotavirus and calicivirus. Waterborne contamination can affect large numbers of individuals [3,9–11]. In the present study, assuming a 50% increase in the population during the tourist season, the number of affected people was estimated to be between 2400 and 2900. The cohort study showed that almost half (46.5%) of the cases consulted a GP. Applying this proportion to the estimated number of cases yielded an estimated 1116–1349 ill individuals who consulted a GP for gastroenteritis, similar to the number of consultations for gastroenteritis actually reported by the GPs (n = 1037). The stool sample analyses revealed the presence of multiple pathogens (rotavirus A, norovirus genogroups I and II and C. coli) as single or co-infections. Except for C. coli, these pathogens have been implicated in numerous waterborne outbreaks, suggesting an influx of sewage into the water supply system [10,12–18]. Rotaviruses are known to be responsible for large epidemics during winter seasons, affecting particularly children aged < 3 years. These epidemics are caused by the faecal route of contamination (personto-person transmission). The high proportion of stools positive for rotavirus (71%) is uncommon in waterborne outbreaks and probably reflects massive contamination [15,16,19]. Different molecular sequences of rotavirus A genotype G1 found in stools and water indicated a human faecal source of infection. In the present study, contamination of water by sewage could be linked with the large amount of sewage from the water purification plant sprayed on to land close to the river and the water catchments. In this outbreak, C. coli, which was identified in 32% of stools, was the second most frequent pathogen. To our knowledge, human C. coli contamination has not been described previously as a causative agent in waterborne outbreaks, unlike C. jejuni [11,20–25], although C. coli has been isolated from water [26]. The main symptoms observed (diarrhoea, abdominal pain, nausea) and the mean duration of the disease (4.5 days) were compatible with a Campylobacter infection. No bloody diarrhoea was reported, although bloody diarrhoea is not constant in Campylobacter infections [27]. The high proportion of individuals who had to take sick leave, and the fact that six individuals were hospitalised, suggested a relative severity of symptoms, which is much more compatible with Campylobacter infection than with virus infections. Among the hospitalised patients, 60% were infected with C. coli. Among patients whose stools were analysed and whose age was known, cases infected only with rotavirus were younger (mean age 49 years; range, 5 months to 95 years) than cases infected only with Campylobacter (mean age 66 years; range 23–88 years) or co-infected with rotavirus plus Campylobacter (mean age 69 years; range 2–91 years). Moreover, the proportion (52%) of medical visits was unusually high compared with the proportion of visits observed during other waterborne outbreaks [13,16,28]. Campylobacter needs specific atmospheric conditions for growth and survival and its detection can be very difficult [29,30]. Because of a rapid physiological transformation to a viable, though non-cultivable, state, it has been suggested that conventional culture methods often fail [31]. Nevertheless, in the present study, the failure to detect C. coli in water, and the detection of noroviruses in stools, could be caused by the small (10 mL) water samples examined. Ha¨nninen et al. [30] proposed that the volume used for detection of suspected pathogens in drinking water should be at least 8–10 L. Groundwater is considered widely to be microbiologically clean, and is usually used for drinking without treatment. However, groundwater can be contaminated by the environment [26,32]. Stanley et al. [33] showed that Campylobacter can occur in groundwater, with hydrological evidence suggesting that the source of contamination in their study was a dairy farm situated within the hydrological catchment area of the groundwater, which was confirmed when identical strains of C. jejuni were isolated from groundwater and the dairy herd. In the present study, Campylobacter was not identified from groundwater, and dairy herds and other domestic animals in farms around the source were not investigated. However, results of bacterial, chemical and physical water analyses suggested strongly that a river contaminated by agricultural run-off from surrounding sheep and cattle grazing meadows was contaminating the source. In addition, the immediate area of the source was insufficiently protected and vulnerable to the environment. Wild birds, rodents, pigs and cows are all hosts for C. coli and could contaminate surface water easily [26,27,34,35]. Campylobacters are sensitive to many environmental factors and to chlorine [33,36,37], but a failure of the chlorination system could have resulted in the presence of a high concentration of C. coli. Waterborne outbreaks are often caused by several factors that act together [18,20,23,38]. In some outbreaks, heavy rainfall has been responsible for increasing agricultural run-off into rivers, although this did not occur before the outbreak in Gourdon. This underlies the need for increased frequency of routine water quality monitoring, chlorination of the water supply according to environmental conditions, and implementation of preventive measures. This outbreak raised the question of the delay in raising the alert. An outbreak can be recognised by: (i) the poor quality of water detected during regular controls; (ii) complaints about water quality; (iii) an incident in the supply and ℠or sewage water system during maintenance work; and (iv) an increased number of cases of gastroenteritis. In most instances, waterborne outbreaks are suspected when there are already a large number of declared cases [3,12,18,28,38,39]. In the present outbreak, the initial warning was given by a GP who notified officials 5–10 days after the assumed date of onset. This highlights the role of GPs in the detection of an unusual event, particularly during the summer, when an increase in acute gastroenteritis is less expected than during winter [40,41]. In addition, a routine water quality monitoring system able to detect contamination of drinking water before the occurrence of illness and complaints about the water quality should be used as an early warning source. The retrospective cohort study of Gourdon residents allowed the scale and severity of the outbreak to be quantified precisely and the hypothesis that tap water was the vehicle of the outbreak to be tested. When provided with a cohort sample representative of the study population, the number of people affected can be estimated. However, as the exposure of the population was high, a relatively large sample size was needed to test the association between the disease and the exposure. Because of misclassification of exposure (unrecognised consumption of tap water), the attributable risk is probably underestimated. In the absence of obvious alternative explanations, water contamination is the most likely common cause of an explosive outbreak of AGE occurring in a large population. Tillett et al. [42] proposed categorising levels of evidence for waterborne outbreaks. A descriptive study suggesting a water-related outbreak, together with the identification of the same pathogen in stools and water, should be sufficient to conclude that an outbreak is associated strongly with water consumption. Faecal indicators of water quality are currently insufficiently sensitive for some pathogens, notably viruses and parasites (e.g., Cryptosporidium). Therefore, in the absence of evidence from an analytical study to demonstrate an association, the link between illness and water can only be classified as probable or possible. However, total coliform counts are good indicators of faecal contamination with other pathogens (e.g., rotaviruses, E. coli, Campylobacter) [43]. In conclusion, the waterborne outbreak described in the present study was most probably caused by faecal contamination of source A by groundwater, resulting from deficiencies in the maintenance of the water distribution system. Corrective measures were implemented based on the results of the investigation. To our knowledge, this is the first documented waterborne outbreak involving human C. coli infection. In-depth environmental investigations to identify the origin of water contamination are important in order to better understand microbial transmission from the environment and to implement appropriate controls and preventive measures. In addition, surveys on vehicles and vectors that transmit Campylobacter between hosts are crucial in order to better understand the epidemiology of these organisms.

What is the date of this event?

282

A large multi-pathogen waterborne community outbreak linked to faecal contamination of a groundwater system, France, 2000

A large waterborne outbreak of infection that occurred during August 2000 in a local community in France was investigated initially via a rapid survey of visits to local physicians. A retrospective cohort study was then conducted on a random cluster sample of residents. Of 709 residents interviewed, 202 (28.5%) were definite cases (at least three liquid stools ℠day or vomiting) and 62 (8.7%) were probable cases (less than three liquid stools ℠day or abdominal pain). Those who had drunk tap water had a threefold increased risk for illness (95% CI 2.4–4.0). The risk increased with the amount of water consumed (chi-square trend: p < 0.0001). Bacteriological analyses of stools were performed for 35 patients and virological analyses for 24 patients. Campylobacter coli, group A rotavirus and norovirus were detected in 31.5%, 71.0% and 21% of samples, respectively. An extensive environmental investigation concluded that a groundwater source to this community had probably been contaminated by agricultural run-off, and a failure in the chlorination system was identified. This is the first documented waterborne outbreak of infection involving human C. coli infections. A better understanding of the factors influencing campylobacter transmission between hosts is required. Keywords Campylobacter coli, epidemiology, gastroenteritis, norovirus, rotavirus, waterborne outbreak INTRODUCTION Contaminated drinking water causes extensive outbreaks of illness because of the large number of people served by water supply facilities. Detection of such outbreaks requires the identification of an increase in illness (usually the rate of gastrointestinal disease) in the exposed population and confirmation that water was the route of transmission [1]. Many countries have routine reporting systems for detecting foodborne and waterborne outbreaks of infection. However, in routine practice it is not easy to detect an increase in clinical cases and to link this increase to waterborne transmission; thus waterborne outbreaks are often unrecognised and underestimated [2]. Microbiological examination of water is more complex than examination of stools, and there is often a failure to detect pathogens in water. Norovirus, Cryptosporidium parvum, Giardia intestinalis and Campylobacter jejuni are the pathogens identified most frequently [3]. The present study describes a large waterborne community outbreak of infection with multiple pathogens, including Campylobacter coli. METHODS AND MATERIALS Background On 23 August 2000, a general practitioner (GP) informed the local district health department that 16 cases of acute gastroenteritis (AGE) had occurred among residents of a holiday camp in the Gourdon community since 21 August. Gourdon, which is located in the south-west of France, has a population of 4888 inhabitants, which increases by 50% in the July– August tourist season. Investigations at the holiday camp concluded that foodborne transmission was very unlikely, but found that tap water was highly contaminated with faecal coliforms. Preliminary information collected from physicians in Gourdon indicated an increase of the number of consultations for AGE. Gourdon tap water is supplied by two underground water facilities: facility A serving 3800 water consumers in Gourdon; and facility B serving 1200 water consumers in Gourdon and the surrounding communities. Water treatment in facility A consists of chlorination, which involves two automatic pumps working in rotation. The water then passes through two semi in-ground reservoir tanks (1100 m3 and 200 m3, respectively) before distribution. Facility B water is mixed with a fraction of facility A water in another 500 m3 semi in-ground reservoir tank before being distributed. This mixing with the treated water from facility A is intended to disinfect the water from facility B. Since the increase in AGE was noted only in Gourdon, contaminated tap water from facility A was suspected to be the origin of the outbreak. Therefore, the population of Gourdon was informed on 25 August to avoid the consumption of tap water or to boil it for 5 min. Survey of gastroenteritis-related medical consultations To confirm the outbreak, the 11 GPs in Gourdon and the local hospital emergency department were requested to provide the number of total daily visits and the number of cases of gastroenteritis cases occurring between 1 August and 11 September 2000. Retrospective cohort study A retrospective cohort study was performed on a sample of Gourdon households between 5 and 15 September 2000. Domestic households were selected randomly from the telephone directory. A definite case was defined as a Gourdon resident who had at least three liquid stools ℠day or vomiting between 1 August and 3 September, and who had been present in Gourdon for ‡ 1 day between 1 and 31 August; if less than three liquid stools ℠day or abdominal pain had occurred, the case was classified as probable. Exposure was defined as the consumption of tap water before 25 August 2000. A standardised telephone questionnaire was completed to obtain demographic data (gender, date of birth), food consumption (commercially prepared dishes, sea-food, raw milk, pastry), tap water consumption (before and after 25 August, daily number of glasses of water, consumption of boiled tap water), symptoms of gastroenteritis (date of onset, number of liquid stools ℠day, vomiting, nausea, abdominal pain, fever), medical care (in- or out-patient) and consequences on daily activities (sick leave, bed confinement) for each member of the selected household. Under the hypothesis that 60% of the population consumed tap water, an estimated 550 people were required for inclusion in the study to identify a relative risk (RR) of ‡ 3, with a 5% a risk and a power of 80%. Based on an average of two individuals per household (INSERM 1999 population census), 300 households were required for inclusion in the survey. The association between AGE and the consumption of tap water was assessed by calculating the risk ratio (RR) and 95% CI, taking into account a household cluster design effect. A dose-effect relationship was evaluated by analysing the RR trend for different categories of daily number of glasses of tap water consumed (chi-square trend). Proportions were compared using the corrected chi-square test, and means were compared using the Student test or Mann and Witney test. Data were analysed with Epi-Info 6.04c (CDC, Atlanta, GA, USA). The gender and age distribution of the cohort population were compared with that of the general population of Gourdon (INSERM 1999 population census) to assess whether it was representative. Under the hypothesis of representation, the total number of cases of gastroenteritis was estimated by applying the attack rate to the total population of Gourdon, including the tourist population (data provided by the Gourdon tourism office). The most likely period for contamination of the water distribution system was based on the epidemic curve and the minimal incubation period of the pathogens involved. Microbiological analyses of stools Stool samples were examined for the presence of Salmonella spp., Shigella spp., Staphylococcus aureus, Campylobacter spp., Yersinia enterocolitica, Escherichia coli, enterococci, rotavirus groups A and C, astrovirus, calicivirus, adenovirus types 40 and 41, enterovirus, hepatitis A virus and Cryptosporidum. Viruses were detected using an enzyme immunoassay for group A rotavirus, astrovirus and adenovirus 40 ℠41, and by RT-PCR for calicivirus, group C rotavirus, enterovirus and hepatitis A virus [4–6]. Campylobacter isolates were sent to the National Reference Centre for Campylobacter for identification and fingerprinting. Species identification was by means of phenotypic tests (API Campy; bioMe´rieux, Lyon, France), with use of PCR to differentiate Campylobacter jejuni from Camplylobacter coli [7]. Random amplified polymorphic DNA fingerprinting was performed with primer 3881, 5¢-AACGCGCAAC. Environmental investigation An environmental and microbiological investigation of the water distribution system was performed. The inspection concerned the water catchments, the water distribution and sewage system, agricultural practices and private households around the sources. Bacteriological (Escherichia coli, enterococci, faecal streptococci, thermotolerant coliforms, sulphite-reducing clostridia) and chemical (chlorine) analyses were performed on tap water and on the water sampled from the two groundwater sources. After 23 August, the water chlorine concentration and the presence of coliform bacteria were monitored daily at several sites of the water system before and after the resort treatment. Examinations for viruses (group A rotavirus, astrovirus, norovirus, enterovirus, hepatitis A virus) and Campylobacter spp. were undertaken on two 10-mL tap water samples collected inhouse on 23 August, and on two 10-mL samples from groundwater sources A and B on 25 August. Viruses were detected by RT-PCR, followed by hybridisation with a specific primer to improve sensitivity, while C. jejuni and C. coli were detected by PCR with specific primers. Group A rotavirus isolates detected in stools and in water were sequenced using Beg 9 and End 9 primers [8] and compared using the Infobiogene GCG software (http://www.infobiogen.fr). Bacteriological (revived 36C, revived 22C, thermotolerant coliforms, streptococci and faecal enterococci), physical (temperature, conductivity) and chemical (oxygen consumption, ammonium, nitrates, phosphorus) parameters from source A and the river were analysed. Trends were compared for an 8-week period between 21 December 2000 and 7 February 2001. RESULTS Epidemiology Between 1 August and 11 September 2000, 7104 Gourdon residents consulted a GP or the local hospital emergency department. Between 24 and 28 August, gastroenteritis accounted for 44% (479 ℠1097) of all medical visits, compared with 6% (126 ℠2185) during the first 15 days of August (Fig. 1). Among the 498 households contacted for the household survey, 198 were excluded (103 did not respond, 49 did not fulfil the inclusion criteria, 43 refused to participate, and three for unknown reasons). The questionnaire was completed for 300 households (709 individuals). Data were missing concerning gender for one individual and age for 11 individuals. Overall, 331 (46.7%; 95% CI 44.5–49.0) of respondents were male and 377 (53.2%; 95% CI 50.9–55.5) were female. The median age was 44 years (range < 1–94 years). The age and gender distribution of the sample was similar to that of the general population. During the outbreak period, 264 of the 709 respondents (attack rate (AR) 37.2%) had been ill, with 202 definite cases (AR 28.5%; 95% CI 24.6– 32.4) and 62 probable cases (AR 8.7%; 95% CI 6.5– 11.0). The AR was higher in females (31.8%) than in males (24.8%), and was highest in children aged < 6 years (42.1%, Table 1). The epidemic curve (Fig. 2) suggests that the outbreak started between 14 and 19 August. The shape of the epidemic curve in the retrospective cohort study was very similar to that based on the number of visits to GPs for gastroenteritis. Patients complained mainly of diarrhoea, abdominal pain and nausea (Table 2). No cases of bloody diarrhoea were notified. The mean duration of symptoms was 4.5 days (median 3, range 1–30 days). Among definite cases, 52% had visited a GP, 42% had taken sick leave for ‡ 1 day, 32% had required bed rest, and 2.9% (six patients) had been hospitalised. In total, 336 (47.4%) respondents had drunk tap water before 25 August, whereas 646 (91.1%) did not drink tap water after 25 August. The risk for illness, taking into account the sample cluster design effect of 1.4, was three-fold greater (95% CI 2.2–4.1) for those who had drunk tap water before 25 August, and increased with the amount of water consumed (chi-square trend: p < 0.0001; Table 3). The risk of gastroenteritis associated with drinking tap water did not change substantially after adjustment for age (RRMH 3.3, 95% CI 2.5–4.4), and there was no modification effect according to age. Food consumption was not associated with illness. Applying the AR of 37% to the total population of Gourdon exposed to the water supply (4888 residents + 2200 tourists), the number of individuals affected by this outbreak was estimated to be 2600 (95% CI 2400–2900), assuming an exposure to tap water and a risk for gastroenteritis among the tourists similar to that among the inhabitants. Microbiological stool analyses Among the 35 stool samples analysed for bacteria, 11 (31.5%) were positive for C. coli. Twenty-four of the 35 samples were also tested for viruses: rotavirus A was detected in 17 (71.0%) and norovirus in five (21.0%) samples, while nine (37.5%) samples were co-infected (four C. coli + rotavirus; one C. coli + norovirus; three rotavirus + norovirus; and one C. coli + rotavirus + norovirus). Of the 11 C. coli isolates, six were characterised as biotype 2, and were identical following RAPD analysis, while the remaining five were of biotype 1. Rotavirus typing by RTPCR identified a single genotype P-(8), G1. Molecular characterisation of the norovirus isolates identified two different patterns, with three isolates being identical and close to the Saragota strain (genogroup I) and two close to the new GGIIb variant (genogroup II). Environmental investigation All of the water samples from source B were negative, regardless of the site of the distribution system tested. Analyses of the water originating from source A, sampled after leaving the water-works on 23 and 25 August, revealed high concentrations of E. coli (2263 / 100 mL), enterococci (415 / 100 mL), total coliforms (90 / 100 mL), faecal streptococci (136 / 100 mL) and sulphite-reducing clostridia (10 / 100 mL), and an absence of chlorination. Rotavirus A genotype G1 was detected in the crude water sampled from source A. The analyses for enterovirus, astrovirus, norovirus, hepatitis A virus and Campylobacter were negative. Source A was situated in a valley containing a river, below a hamlet of 15 dwellings, including farms breeding cattle, pork and sheep. The source was not supposed to be supplied by the river. The river resulted from the convergence of about ten different streams, with one originating close to a slaughterhouse, and was exposed to agricultural run-off and manure effluent; sheep and cattle grazed in meadows bordering the river, and a large amount of sewage from water treatment facilities was sprayed on to land close to the river. Cowsheds were close to the water catchment area. The immediate protected area around source A was too small and littered with building rubbish. Air vents without protection in the door made the source accessible to animals and insects. Bacteriological analyses of the water samples from source A and the river revealed that the contamination was of both human and animal origin, with a higher concentration in the river. The levels and trends of the chemical parameters of source A were similar to those of the river, strongly suggesting that the river was connected to source A (Fig. 3). A retrospective review of routine data collected between 1991 and 2000 showed irregularities in the maintenance of the water distribution system and the water treatment plant, particularly during the month of August. In the hamlet, the sewage system of the 15 households did not comply with sanitary rules. All but three ejected waste either into a cesspool or superficially without treatment. Three regions of gene 9 of the rotavirus detected in stools and water were compared. Sequencing showed that the strains were closely related but not identical. Assuming a likely onset of the outbreak between 14 and 19 August, and taking into account a minimum incubation period of 2 days for rotavirus A, norovirus and C. coli, the contamination of the water distribution system probably occurred between 12 and 17 August. Contamination probably persisted until 28 August, as intensive chlorination, beginning on 25 August, was effective only from 28 August onwards (Fig. 2). DISCUSSION The present descriptive and retrospective cohort studies confirmed that the outbreak of infection in the community of Gourdon was related to the consumption of tap water and affected 2600 individuals. The outbreak involved multiple pathogens, with group A rotavirus and C. coli being the most frequent, followed by norovirus. Results of environmental investigations suggested that water contamination was related to the contamination of the source, combined with a failure of the water chlorination system. The association with tap water consumption, the dose-effect relationship, and the fact that multiple pathogens were involved, were all strongly in favour of tap water as the origin of infection. Moreover, the number of cases decreased dramatically after the water restriction advice and the implementation of control measures on the water supply. No food products explored in the study were associated with illness. An AR of 14.5% and 31% was observed among adults and children aged < 6 years, respectively, who reported that they had not consumed tap water. This relatively high rate could be explained by several phenomena. Many individuals may have consumed tap water without being aware of it, e.g., in drinks or ice cubes consumed outside of the home, via raw fruits and vegetables, dishes prepared with unboiled water, or while cleaning their teeth. It can be hypothesised that people who became ill without being aware of having consumed tap water were exposed to small amounts of water and had been subjected to a small infective dose. In addition, children, more than adults, may have been contaminated through person-to-person transmission (e.g., family contact), which is the most common route of transmission for rotavirus and calicivirus. Waterborne contamination can affect large numbers of individuals [3,9–11]. In the present study, assuming a 50% increase in the population during the tourist season, the number of affected people was estimated to be between 2400 and 2900. The cohort study showed that almost half (46.5%) of the cases consulted a GP. Applying this proportion to the estimated number of cases yielded an estimated 1116–1349 ill individuals who consulted a GP for gastroenteritis, similar to the number of consultations for gastroenteritis actually reported by the GPs (n = 1037). The stool sample analyses revealed the presence of multiple pathogens (rotavirus A, norovirus genogroups I and II and C. coli) as single or co-infections. Except for C. coli, these pathogens have been implicated in numerous waterborne outbreaks, suggesting an influx of sewage into the water supply system [10,12–18]. Rotaviruses are known to be responsible for large epidemics during winter seasons, affecting particularly children aged < 3 years. These epidemics are caused by the faecal route of contamination (personto-person transmission). The high proportion of stools positive for rotavirus (71%) is uncommon in waterborne outbreaks and probably reflects massive contamination [15,16,19]. Different molecular sequences of rotavirus A genotype G1 found in stools and water indicated a human faecal source of infection. In the present study, contamination of water by sewage could be linked with the large amount of sewage from the water purification plant sprayed on to land close to the river and the water catchments. In this outbreak, C. coli, which was identified in 32% of stools, was the second most frequent pathogen. To our knowledge, human C. coli contamination has not been described previously as a causative agent in waterborne outbreaks, unlike C. jejuni [11,20–25], although C. coli has been isolated from water [26]. The main symptoms observed (diarrhoea, abdominal pain, nausea) and the mean duration of the disease (4.5 days) were compatible with a Campylobacter infection. No bloody diarrhoea was reported, although bloody diarrhoea is not constant in Campylobacter infections [27]. The high proportion of individuals who had to take sick leave, and the fact that six individuals were hospitalised, suggested a relative severity of symptoms, which is much more compatible with Campylobacter infection than with virus infections. Among the hospitalised patients, 60% were infected with C. coli. Among patients whose stools were analysed and whose age was known, cases infected only with rotavirus were younger (mean age 49 years; range, 5 months to 95 years) than cases infected only with Campylobacter (mean age 66 years; range 23–88 years) or co-infected with rotavirus plus Campylobacter (mean age 69 years; range 2–91 years). Moreover, the proportion (52%) of medical visits was unusually high compared with the proportion of visits observed during other waterborne outbreaks [13,16,28]. Campylobacter needs specific atmospheric conditions for growth and survival and its detection can be very difficult [29,30]. Because of a rapid physiological transformation to a viable, though non-cultivable, state, it has been suggested that conventional culture methods often fail [31]. Nevertheless, in the present study, the failure to detect C. coli in water, and the detection of noroviruses in stools, could be caused by the small (10 mL) water samples examined. Ha¨nninen et al. [30] proposed that the volume used for detection of suspected pathogens in drinking water should be at least 8–10 L. Groundwater is considered widely to be microbiologically clean, and is usually used for drinking without treatment. However, groundwater can be contaminated by the environment [26,32]. Stanley et al. [33] showed that Campylobacter can occur in groundwater, with hydrological evidence suggesting that the source of contamination in their study was a dairy farm situated within the hydrological catchment area of the groundwater, which was confirmed when identical strains of C. jejuni were isolated from groundwater and the dairy herd. In the present study, Campylobacter was not identified from groundwater, and dairy herds and other domestic animals in farms around the source were not investigated. However, results of bacterial, chemical and physical water analyses suggested strongly that a river contaminated by agricultural run-off from surrounding sheep and cattle grazing meadows was contaminating the source. In addition, the immediate area of the source was insufficiently protected and vulnerable to the environment. Wild birds, rodents, pigs and cows are all hosts for C. coli and could contaminate surface water easily [26,27,34,35]. Campylobacters are sensitive to many environmental factors and to chlorine [33,36,37], but a failure of the chlorination system could have resulted in the presence of a high concentration of C. coli. Waterborne outbreaks are often caused by several factors that act together [18,20,23,38]. In some outbreaks, heavy rainfall has been responsible for increasing agricultural run-off into rivers, although this did not occur before the outbreak in Gourdon. This underlies the need for increased frequency of routine water quality monitoring, chlorination of the water supply according to environmental conditions, and implementation of preventive measures. This outbreak raised the question of the delay in raising the alert. An outbreak can be recognised by: (i) the poor quality of water detected during regular controls; (ii) complaints about water quality; (iii) an incident in the supply and ℠or sewage water system during maintenance work; and (iv) an increased number of cases of gastroenteritis. In most instances, waterborne outbreaks are suspected when there are already a large number of declared cases [3,12,18,28,38,39]. In the present outbreak, the initial warning was given by a GP who notified officials 5–10 days after the assumed date of onset. This highlights the role of GPs in the detection of an unusual event, particularly during the summer, when an increase in acute gastroenteritis is less expected than during winter [40,41]. In addition, a routine water quality monitoring system able to detect contamination of drinking water before the occurrence of illness and complaints about the water quality should be used as an early warning source. The retrospective cohort study of Gourdon residents allowed the scale and severity of the outbreak to be quantified precisely and the hypothesis that tap water was the vehicle of the outbreak to be tested. When provided with a cohort sample representative of the study population, the number of people affected can be estimated. However, as the exposure of the population was high, a relatively large sample size was needed to test the association between the disease and the exposure. Because of misclassification of exposure (unrecognised consumption of tap water), the attributable risk is probably underestimated. In the absence of obvious alternative explanations, water contamination is the most likely common cause of an explosive outbreak of AGE occurring in a large population. Tillett et al. [42] proposed categorising levels of evidence for waterborne outbreaks. A descriptive study suggesting a water-related outbreak, together with the identification of the same pathogen in stools and water, should be sufficient to conclude that an outbreak is associated strongly with water consumption. Faecal indicators of water quality are currently insufficiently sensitive for some pathogens, notably viruses and parasites (e.g., Cryptosporidium). Therefore, in the absence of evidence from an analytical study to demonstrate an association, the link between illness and water can only be classified as probable or possible. However, total coliform counts are good indicators of faecal contamination with other pathogens (e.g., rotaviruses, E. coli, Campylobacter) [43]. In conclusion, the waterborne outbreak described in the present study was most probably caused by faecal contamination of source A by groundwater, resulting from deficiencies in the maintenance of the water distribution system. Corrective measures were implemented based on the results of the investigation. To our knowledge, this is the first documented waterborne outbreak involving human C. coli infection. In-depth environmental investigations to identify the origin of water contamination are important in order to better understand microbial transmission from the environment and to implement appropriate controls and preventive measures. In addition, surveys on vehicles and vectors that transmit Campylobacter between hosts are crucial in order to better understand the epidemiology of these organisms.

How long was the event?

283

A large multi-pathogen waterborne community outbreak linked to faecal contamination of a groundwater system, France, 2000

A large waterborne outbreak of infection that occurred during August 2000 in a local community in France was investigated initially via a rapid survey of visits to local physicians. A retrospective cohort study was then conducted on a random cluster sample of residents. Of 709 residents interviewed, 202 (28.5%) were definite cases (at least three liquid stools ℠day or vomiting) and 62 (8.7%) were probable cases (less than three liquid stools ℠day or abdominal pain). Those who had drunk tap water had a threefold increased risk for illness (95% CI 2.4–4.0). The risk increased with the amount of water consumed (chi-square trend: p < 0.0001). Bacteriological analyses of stools were performed for 35 patients and virological analyses for 24 patients. Campylobacter coli, group A rotavirus and norovirus were detected in 31.5%, 71.0% and 21% of samples, respectively. An extensive environmental investigation concluded that a groundwater source to this community had probably been contaminated by agricultural run-off, and a failure in the chlorination system was identified. This is the first documented waterborne outbreak of infection involving human C. coli infections. A better understanding of the factors influencing campylobacter transmission between hosts is required. Keywords Campylobacter coli, epidemiology, gastroenteritis, norovirus, rotavirus, waterborne outbreak INTRODUCTION Contaminated drinking water causes extensive outbreaks of illness because of the large number of people served by water supply facilities. Detection of such outbreaks requires the identification of an increase in illness (usually the rate of gastrointestinal disease) in the exposed population and confirmation that water was the route of transmission [1]. Many countries have routine reporting systems for detecting foodborne and waterborne outbreaks of infection. However, in routine practice it is not easy to detect an increase in clinical cases and to link this increase to waterborne transmission; thus waterborne outbreaks are often unrecognised and underestimated [2]. Microbiological examination of water is more complex than examination of stools, and there is often a failure to detect pathogens in water. Norovirus, Cryptosporidium parvum, Giardia intestinalis and Campylobacter jejuni are the pathogens identified most frequently [3]. The present study describes a large waterborne community outbreak of infection with multiple pathogens, including Campylobacter coli. METHODS AND MATERIALS Background On 23 August 2000, a general practitioner (GP) informed the local district health department that 16 cases of acute gastroenteritis (AGE) had occurred among residents of a holiday camp in the Gourdon community since 21 August. Gourdon, which is located in the south-west of France, has a population of 4888 inhabitants, which increases by 50% in the July– August tourist season. Investigations at the holiday camp concluded that foodborne transmission was very unlikely, but found that tap water was highly contaminated with faecal coliforms. Preliminary information collected from physicians in Gourdon indicated an increase of the number of consultations for AGE. Gourdon tap water is supplied by two underground water facilities: facility A serving 3800 water consumers in Gourdon; and facility B serving 1200 water consumers in Gourdon and the surrounding communities. Water treatment in facility A consists of chlorination, which involves two automatic pumps working in rotation. The water then passes through two semi in-ground reservoir tanks (1100 m3 and 200 m3, respectively) before distribution. Facility B water is mixed with a fraction of facility A water in another 500 m3 semi in-ground reservoir tank before being distributed. This mixing with the treated water from facility A is intended to disinfect the water from facility B. Since the increase in AGE was noted only in Gourdon, contaminated tap water from facility A was suspected to be the origin of the outbreak. Therefore, the population of Gourdon was informed on 25 August to avoid the consumption of tap water or to boil it for 5 min. Survey of gastroenteritis-related medical consultations To confirm the outbreak, the 11 GPs in Gourdon and the local hospital emergency department were requested to provide the number of total daily visits and the number of cases of gastroenteritis cases occurring between 1 August and 11 September 2000. Retrospective cohort study A retrospective cohort study was performed on a sample of Gourdon households between 5 and 15 September 2000. Domestic households were selected randomly from the telephone directory. A definite case was defined as a Gourdon resident who had at least three liquid stools ℠day or vomiting between 1 August and 3 September, and who had been present in Gourdon for ‡ 1 day between 1 and 31 August; if less than three liquid stools ℠day or abdominal pain had occurred, the case was classified as probable. Exposure was defined as the consumption of tap water before 25 August 2000. A standardised telephone questionnaire was completed to obtain demographic data (gender, date of birth), food consumption (commercially prepared dishes, sea-food, raw milk, pastry), tap water consumption (before and after 25 August, daily number of glasses of water, consumption of boiled tap water), symptoms of gastroenteritis (date of onset, number of liquid stools ℠day, vomiting, nausea, abdominal pain, fever), medical care (in- or out-patient) and consequences on daily activities (sick leave, bed confinement) for each member of the selected household. Under the hypothesis that 60% of the population consumed tap water, an estimated 550 people were required for inclusion in the study to identify a relative risk (RR) of ‡ 3, with a 5% a risk and a power of 80%. Based on an average of two individuals per household (INSERM 1999 population census), 300 households were required for inclusion in the survey. The association between AGE and the consumption of tap water was assessed by calculating the risk ratio (RR) and 95% CI, taking into account a household cluster design effect. A dose-effect relationship was evaluated by analysing the RR trend for different categories of daily number of glasses of tap water consumed (chi-square trend). Proportions were compared using the corrected chi-square test, and means were compared using the Student test or Mann and Witney test. Data were analysed with Epi-Info 6.04c (CDC, Atlanta, GA, USA). The gender and age distribution of the cohort population were compared with that of the general population of Gourdon (INSERM 1999 population census) to assess whether it was representative. Under the hypothesis of representation, the total number of cases of gastroenteritis was estimated by applying the attack rate to the total population of Gourdon, including the tourist population (data provided by the Gourdon tourism office). The most likely period for contamination of the water distribution system was based on the epidemic curve and the minimal incubation period of the pathogens involved. Microbiological analyses of stools Stool samples were examined for the presence of Salmonella spp., Shigella spp., Staphylococcus aureus, Campylobacter spp., Yersinia enterocolitica, Escherichia coli, enterococci, rotavirus groups A and C, astrovirus, calicivirus, adenovirus types 40 and 41, enterovirus, hepatitis A virus and Cryptosporidum. Viruses were detected using an enzyme immunoassay for group A rotavirus, astrovirus and adenovirus 40 ℠41, and by RT-PCR for calicivirus, group C rotavirus, enterovirus and hepatitis A virus [4–6]. Campylobacter isolates were sent to the National Reference Centre for Campylobacter for identification and fingerprinting. Species identification was by means of phenotypic tests (API Campy; bioMe´rieux, Lyon, France), with use of PCR to differentiate Campylobacter jejuni from Camplylobacter coli [7]. Random amplified polymorphic DNA fingerprinting was performed with primer 3881, 5¢-AACGCGCAAC. Environmental investigation An environmental and microbiological investigation of the water distribution system was performed. The inspection concerned the water catchments, the water distribution and sewage system, agricultural practices and private households around the sources. Bacteriological (Escherichia coli, enterococci, faecal streptococci, thermotolerant coliforms, sulphite-reducing clostridia) and chemical (chlorine) analyses were performed on tap water and on the water sampled from the two groundwater sources. After 23 August, the water chlorine concentration and the presence of coliform bacteria were monitored daily at several sites of the water system before and after the resort treatment. Examinations for viruses (group A rotavirus, astrovirus, norovirus, enterovirus, hepatitis A virus) and Campylobacter spp. were undertaken on two 10-mL tap water samples collected inhouse on 23 August, and on two 10-mL samples from groundwater sources A and B on 25 August. Viruses were detected by RT-PCR, followed by hybridisation with a specific primer to improve sensitivity, while C. jejuni and C. coli were detected by PCR with specific primers. Group A rotavirus isolates detected in stools and in water were sequenced using Beg 9 and End 9 primers [8] and compared using the Infobiogene GCG software (http://www.infobiogen.fr). Bacteriological (revived 36C, revived 22C, thermotolerant coliforms, streptococci and faecal enterococci), physical (temperature, conductivity) and chemical (oxygen consumption, ammonium, nitrates, phosphorus) parameters from source A and the river were analysed. Trends were compared for an 8-week period between 21 December 2000 and 7 February 2001. RESULTS Epidemiology Between 1 August and 11 September 2000, 7104 Gourdon residents consulted a GP or the local hospital emergency department. Between 24 and 28 August, gastroenteritis accounted for 44% (479 ℠1097) of all medical visits, compared with 6% (126 ℠2185) during the first 15 days of August (Fig. 1). Among the 498 households contacted for the household survey, 198 were excluded (103 did not respond, 49 did not fulfil the inclusion criteria, 43 refused to participate, and three for unknown reasons). The questionnaire was completed for 300 households (709 individuals). Data were missing concerning gender for one individual and age for 11 individuals. Overall, 331 (46.7%; 95% CI 44.5–49.0) of respondents were male and 377 (53.2%; 95% CI 50.9–55.5) were female. The median age was 44 years (range < 1–94 years). The age and gender distribution of the sample was similar to that of the general population. During the outbreak period, 264 of the 709 respondents (attack rate (AR) 37.2%) had been ill, with 202 definite cases (AR 28.5%; 95% CI 24.6– 32.4) and 62 probable cases (AR 8.7%; 95% CI 6.5– 11.0). The AR was higher in females (31.8%) than in males (24.8%), and was highest in children aged < 6 years (42.1%, Table 1). The epidemic curve (Fig. 2) suggests that the outbreak started between 14 and 19 August. The shape of the epidemic curve in the retrospective cohort study was very similar to that based on the number of visits to GPs for gastroenteritis. Patients complained mainly of diarrhoea, abdominal pain and nausea (Table 2). No cases of bloody diarrhoea were notified. The mean duration of symptoms was 4.5 days (median 3, range 1–30 days). Among definite cases, 52% had visited a GP, 42% had taken sick leave for ‡ 1 day, 32% had required bed rest, and 2.9% (six patients) had been hospitalised. In total, 336 (47.4%) respondents had drunk tap water before 25 August, whereas 646 (91.1%) did not drink tap water after 25 August. The risk for illness, taking into account the sample cluster design effect of 1.4, was three-fold greater (95% CI 2.2–4.1) for those who had drunk tap water before 25 August, and increased with the amount of water consumed (chi-square trend: p < 0.0001; Table 3). The risk of gastroenteritis associated with drinking tap water did not change substantially after adjustment for age (RRMH 3.3, 95% CI 2.5–4.4), and there was no modification effect according to age. Food consumption was not associated with illness. Applying the AR of 37% to the total population of Gourdon exposed to the water supply (4888 residents + 2200 tourists), the number of individuals affected by this outbreak was estimated to be 2600 (95% CI 2400–2900), assuming an exposure to tap water and a risk for gastroenteritis among the tourists similar to that among the inhabitants. Microbiological stool analyses Among the 35 stool samples analysed for bacteria, 11 (31.5%) were positive for C. coli. Twenty-four of the 35 samples were also tested for viruses: rotavirus A was detected in 17 (71.0%) and norovirus in five (21.0%) samples, while nine (37.5%) samples were co-infected (four C. coli + rotavirus; one C. coli + norovirus; three rotavirus + norovirus; and one C. coli + rotavirus + norovirus). Of the 11 C. coli isolates, six were characterised as biotype 2, and were identical following RAPD analysis, while the remaining five were of biotype 1. Rotavirus typing by RTPCR identified a single genotype P-(8), G1. Molecular characterisation of the norovirus isolates identified two different patterns, with three isolates being identical and close to the Saragota strain (genogroup I) and two close to the new GGIIb variant (genogroup II). Environmental investigation All of the water samples from source B were negative, regardless of the site of the distribution system tested. Analyses of the water originating from source A, sampled after leaving the water-works on 23 and 25 August, revealed high concentrations of E. coli (2263 / 100 mL), enterococci (415 / 100 mL), total coliforms (90 / 100 mL), faecal streptococci (136 / 100 mL) and sulphite-reducing clostridia (10 / 100 mL), and an absence of chlorination. Rotavirus A genotype G1 was detected in the crude water sampled from source A. The analyses for enterovirus, astrovirus, norovirus, hepatitis A virus and Campylobacter were negative. Source A was situated in a valley containing a river, below a hamlet of 15 dwellings, including farms breeding cattle, pork and sheep. The source was not supposed to be supplied by the river. The river resulted from the convergence of about ten different streams, with one originating close to a slaughterhouse, and was exposed to agricultural run-off and manure effluent; sheep and cattle grazed in meadows bordering the river, and a large amount of sewage from water treatment facilities was sprayed on to land close to the river. Cowsheds were close to the water catchment area. The immediate protected area around source A was too small and littered with building rubbish. Air vents without protection in the door made the source accessible to animals and insects. Bacteriological analyses of the water samples from source A and the river revealed that the contamination was of both human and animal origin, with a higher concentration in the river. The levels and trends of the chemical parameters of source A were similar to those of the river, strongly suggesting that the river was connected to source A (Fig. 3). A retrospective review of routine data collected between 1991 and 2000 showed irregularities in the maintenance of the water distribution system and the water treatment plant, particularly during the month of August. In the hamlet, the sewage system of the 15 households did not comply with sanitary rules. All but three ejected waste either into a cesspool or superficially without treatment. Three regions of gene 9 of the rotavirus detected in stools and water were compared. Sequencing showed that the strains were closely related but not identical. Assuming a likely onset of the outbreak between 14 and 19 August, and taking into account a minimum incubation period of 2 days for rotavirus A, norovirus and C. coli, the contamination of the water distribution system probably occurred between 12 and 17 August. Contamination probably persisted until 28 August, as intensive chlorination, beginning on 25 August, was effective only from 28 August onwards (Fig. 2). DISCUSSION The present descriptive and retrospective cohort studies confirmed that the outbreak of infection in the community of Gourdon was related to the consumption of tap water and affected 2600 individuals. The outbreak involved multiple pathogens, with group A rotavirus and C. coli being the most frequent, followed by norovirus. Results of environmental investigations suggested that water contamination was related to the contamination of the source, combined with a failure of the water chlorination system. The association with tap water consumption, the dose-effect relationship, and the fact that multiple pathogens were involved, were all strongly in favour of tap water as the origin of infection. Moreover, the number of cases decreased dramatically after the water restriction advice and the implementation of control measures on the water supply. No food products explored in the study were associated with illness. An AR of 14.5% and 31% was observed among adults and children aged < 6 years, respectively, who reported that they had not consumed tap water. This relatively high rate could be explained by several phenomena. Many individuals may have consumed tap water without being aware of it, e.g., in drinks or ice cubes consumed outside of the home, via raw fruits and vegetables, dishes prepared with unboiled water, or while cleaning their teeth. It can be hypothesised that people who became ill without being aware of having consumed tap water were exposed to small amounts of water and had been subjected to a small infective dose. In addition, children, more than adults, may have been contaminated through person-to-person transmission (e.g., family contact), which is the most common route of transmission for rotavirus and calicivirus. Waterborne contamination can affect large numbers of individuals [3,9–11]. In the present study, assuming a 50% increase in the population during the tourist season, the number of affected people was estimated to be between 2400 and 2900. The cohort study showed that almost half (46.5%) of the cases consulted a GP. Applying this proportion to the estimated number of cases yielded an estimated 1116–1349 ill individuals who consulted a GP for gastroenteritis, similar to the number of consultations for gastroenteritis actually reported by the GPs (n = 1037). The stool sample analyses revealed the presence of multiple pathogens (rotavirus A, norovirus genogroups I and II and C. coli) as single or co-infections. Except for C. coli, these pathogens have been implicated in numerous waterborne outbreaks, suggesting an influx of sewage into the water supply system [10,12–18]. Rotaviruses are known to be responsible for large epidemics during winter seasons, affecting particularly children aged < 3 years. These epidemics are caused by the faecal route of contamination (personto-person transmission). The high proportion of stools positive for rotavirus (71%) is uncommon in waterborne outbreaks and probably reflects massive contamination [15,16,19]. Different molecular sequences of rotavirus A genotype G1 found in stools and water indicated a human faecal source of infection. In the present study, contamination of water by sewage could be linked with the large amount of sewage from the water purification plant sprayed on to land close to the river and the water catchments. In this outbreak, C. coli, which was identified in 32% of stools, was the second most frequent pathogen. To our knowledge, human C. coli contamination has not been described previously as a causative agent in waterborne outbreaks, unlike C. jejuni [11,20–25], although C. coli has been isolated from water [26]. The main symptoms observed (diarrhoea, abdominal pain, nausea) and the mean duration of the disease (4.5 days) were compatible with a Campylobacter infection. No bloody diarrhoea was reported, although bloody diarrhoea is not constant in Campylobacter infections [27]. The high proportion of individuals who had to take sick leave, and the fact that six individuals were hospitalised, suggested a relative severity of symptoms, which is much more compatible with Campylobacter infection than with virus infections. Among the hospitalised patients, 60% were infected with C. coli. Among patients whose stools were analysed and whose age was known, cases infected only with rotavirus were younger (mean age 49 years; range, 5 months to 95 years) than cases infected only with Campylobacter (mean age 66 years; range 23–88 years) or co-infected with rotavirus plus Campylobacter (mean age 69 years; range 2–91 years). Moreover, the proportion (52%) of medical visits was unusually high compared with the proportion of visits observed during other waterborne outbreaks [13,16,28]. Campylobacter needs specific atmospheric conditions for growth and survival and its detection can be very difficult [29,30]. Because of a rapid physiological transformation to a viable, though non-cultivable, state, it has been suggested that conventional culture methods often fail [31]. Nevertheless, in the present study, the failure to detect C. coli in water, and the detection of noroviruses in stools, could be caused by the small (10 mL) water samples examined. Ha¨nninen et al. [30] proposed that the volume used for detection of suspected pathogens in drinking water should be at least 8–10 L. Groundwater is considered widely to be microbiologically clean, and is usually used for drinking without treatment. However, groundwater can be contaminated by the environment [26,32]. Stanley et al. [33] showed that Campylobacter can occur in groundwater, with hydrological evidence suggesting that the source of contamination in their study was a dairy farm situated within the hydrological catchment area of the groundwater, which was confirmed when identical strains of C. jejuni were isolated from groundwater and the dairy herd. In the present study, Campylobacter was not identified from groundwater, and dairy herds and other domestic animals in farms around the source were not investigated. However, results of bacterial, chemical and physical water analyses suggested strongly that a river contaminated by agricultural run-off from surrounding sheep and cattle grazing meadows was contaminating the source. In addition, the immediate area of the source was insufficiently protected and vulnerable to the environment. Wild birds, rodents, pigs and cows are all hosts for C. coli and could contaminate surface water easily [26,27,34,35]. Campylobacters are sensitive to many environmental factors and to chlorine [33,36,37], but a failure of the chlorination system could have resulted in the presence of a high concentration of C. coli. Waterborne outbreaks are often caused by several factors that act together [18,20,23,38]. In some outbreaks, heavy rainfall has been responsible for increasing agricultural run-off into rivers, although this did not occur before the outbreak in Gourdon. This underlies the need for increased frequency of routine water quality monitoring, chlorination of the water supply according to environmental conditions, and implementation of preventive measures. This outbreak raised the question of the delay in raising the alert. An outbreak can be recognised by: (i) the poor quality of water detected during regular controls; (ii) complaints about water quality; (iii) an incident in the supply and ℠or sewage water system during maintenance work; and (iv) an increased number of cases of gastroenteritis. In most instances, waterborne outbreaks are suspected when there are already a large number of declared cases [3,12,18,28,38,39]. In the present outbreak, the initial warning was given by a GP who notified officials 5–10 days after the assumed date of onset. This highlights the role of GPs in the detection of an unusual event, particularly during the summer, when an increase in acute gastroenteritis is less expected than during winter [40,41]. In addition, a routine water quality monitoring system able to detect contamination of drinking water before the occurrence of illness and complaints about the water quality should be used as an early warning source. The retrospective cohort study of Gourdon residents allowed the scale and severity of the outbreak to be quantified precisely and the hypothesis that tap water was the vehicle of the outbreak to be tested. When provided with a cohort sample representative of the study population, the number of people affected can be estimated. However, as the exposure of the population was high, a relatively large sample size was needed to test the association between the disease and the exposure. Because of misclassification of exposure (unrecognised consumption of tap water), the attributable risk is probably underestimated. In the absence of obvious alternative explanations, water contamination is the most likely common cause of an explosive outbreak of AGE occurring in a large population. Tillett et al. [42] proposed categorising levels of evidence for waterborne outbreaks. A descriptive study suggesting a water-related outbreak, together with the identification of the same pathogen in stools and water, should be sufficient to conclude that an outbreak is associated strongly with water consumption. Faecal indicators of water quality are currently insufficiently sensitive for some pathogens, notably viruses and parasites (e.g., Cryptosporidium). Therefore, in the absence of evidence from an analytical study to demonstrate an association, the link between illness and water can only be classified as probable or possible. However, total coliform counts are good indicators of faecal contamination with other pathogens (e.g., rotaviruses, E. coli, Campylobacter) [43]. In conclusion, the waterborne outbreak described in the present study was most probably caused by faecal contamination of source A by groundwater, resulting from deficiencies in the maintenance of the water distribution system. Corrective measures were implemented based on the results of the investigation. To our knowledge, this is the first documented waterborne outbreak involving human C. coli infection. In-depth environmental investigations to identify the origin of water contamination are important in order to better understand microbial transmission from the environment and to implement appropriate controls and preventive measures. In addition, surveys on vehicles and vectors that transmit Campylobacter between hosts are crucial in order to better understand the epidemiology of these organisms.

How long did the event last?

284

A large multi-pathogen waterborne community outbreak linked to faecal contamination of a groundwater system, France, 2000

A large waterborne outbreak of infection that occurred during August 2000 in a local community in France was investigated initially via a rapid survey of visits to local physicians. A retrospective cohort study was then conducted on a random cluster sample of residents. Of 709 residents interviewed, 202 (28.5%) were definite cases (at least three liquid stools ℠day or vomiting) and 62 (8.7%) were probable cases (less than three liquid stools ℠day or abdominal pain). Those who had drunk tap water had a threefold increased risk for illness (95% CI 2.4–4.0). The risk increased with the amount of water consumed (chi-square trend: p < 0.0001). Bacteriological analyses of stools were performed for 35 patients and virological analyses for 24 patients. Campylobacter coli, group A rotavirus and norovirus were detected in 31.5%, 71.0% and 21% of samples, respectively. An extensive environmental investigation concluded that a groundwater source to this community had probably been contaminated by agricultural run-off, and a failure in the chlorination system was identified. This is the first documented waterborne outbreak of infection involving human C. coli infections. A better understanding of the factors influencing campylobacter transmission between hosts is required. Keywords Campylobacter coli, epidemiology, gastroenteritis, norovirus, rotavirus, waterborne outbreak INTRODUCTION Contaminated drinking water causes extensive outbreaks of illness because of the large number of people served by water supply facilities. Detection of such outbreaks requires the identification of an increase in illness (usually the rate of gastrointestinal disease) in the exposed population and confirmation that water was the route of transmission [1]. Many countries have routine reporting systems for detecting foodborne and waterborne outbreaks of infection. However, in routine practice it is not easy to detect an increase in clinical cases and to link this increase to waterborne transmission; thus waterborne outbreaks are often unrecognised and underestimated [2]. Microbiological examination of water is more complex than examination of stools, and there is often a failure to detect pathogens in water. Norovirus, Cryptosporidium parvum, Giardia intestinalis and Campylobacter jejuni are the pathogens identified most frequently [3]. The present study describes a large waterborne community outbreak of infection with multiple pathogens, including Campylobacter coli. METHODS AND MATERIALS Background On 23 August 2000, a general practitioner (GP) informed the local district health department that 16 cases of acute gastroenteritis (AGE) had occurred among residents of a holiday camp in the Gourdon community since 21 August. Gourdon, which is located in the south-west of France, has a population of 4888 inhabitants, which increases by 50% in the July– August tourist season. Investigations at the holiday camp concluded that foodborne transmission was very unlikely, but found that tap water was highly contaminated with faecal coliforms. Preliminary information collected from physicians in Gourdon indicated an increase of the number of consultations for AGE. Gourdon tap water is supplied by two underground water facilities: facility A serving 3800 water consumers in Gourdon; and facility B serving 1200 water consumers in Gourdon and the surrounding communities. Water treatment in facility A consists of chlorination, which involves two automatic pumps working in rotation. The water then passes through two semi in-ground reservoir tanks (1100 m3 and 200 m3, respectively) before distribution. Facility B water is mixed with a fraction of facility A water in another 500 m3 semi in-ground reservoir tank before being distributed. This mixing with the treated water from facility A is intended to disinfect the water from facility B. Since the increase in AGE was noted only in Gourdon, contaminated tap water from facility A was suspected to be the origin of the outbreak. Therefore, the population of Gourdon was informed on 25 August to avoid the consumption of tap water or to boil it for 5 min. Survey of gastroenteritis-related medical consultations To confirm the outbreak, the 11 GPs in Gourdon and the local hospital emergency department were requested to provide the number of total daily visits and the number of cases of gastroenteritis cases occurring between 1 August and 11 September 2000. Retrospective cohort study A retrospective cohort study was performed on a sample of Gourdon households between 5 and 15 September 2000. Domestic households were selected randomly from the telephone directory. A definite case was defined as a Gourdon resident who had at least three liquid stools ℠day or vomiting between 1 August and 3 September, and who had been present in Gourdon for ‡ 1 day between 1 and 31 August; if less than three liquid stools ℠day or abdominal pain had occurred, the case was classified as probable. Exposure was defined as the consumption of tap water before 25 August 2000. A standardised telephone questionnaire was completed to obtain demographic data (gender, date of birth), food consumption (commercially prepared dishes, sea-food, raw milk, pastry), tap water consumption (before and after 25 August, daily number of glasses of water, consumption of boiled tap water), symptoms of gastroenteritis (date of onset, number of liquid stools ℠day, vomiting, nausea, abdominal pain, fever), medical care (in- or out-patient) and consequences on daily activities (sick leave, bed confinement) for each member of the selected household. Under the hypothesis that 60% of the population consumed tap water, an estimated 550 people were required for inclusion in the study to identify a relative risk (RR) of ‡ 3, with a 5% a risk and a power of 80%. Based on an average of two individuals per household (INSERM 1999 population census), 300 households were required for inclusion in the survey. The association between AGE and the consumption of tap water was assessed by calculating the risk ratio (RR) and 95% CI, taking into account a household cluster design effect. A dose-effect relationship was evaluated by analysing the RR trend for different categories of daily number of glasses of tap water consumed (chi-square trend). Proportions were compared using the corrected chi-square test, and means were compared using the Student test or Mann and Witney test. Data were analysed with Epi-Info 6.04c (CDC, Atlanta, GA, USA). The gender and age distribution of the cohort population were compared with that of the general population of Gourdon (INSERM 1999 population census) to assess whether it was representative. Under the hypothesis of representation, the total number of cases of gastroenteritis was estimated by applying the attack rate to the total population of Gourdon, including the tourist population (data provided by the Gourdon tourism office). The most likely period for contamination of the water distribution system was based on the epidemic curve and the minimal incubation period of the pathogens involved. Microbiological analyses of stools Stool samples were examined for the presence of Salmonella spp., Shigella spp., Staphylococcus aureus, Campylobacter spp., Yersinia enterocolitica, Escherichia coli, enterococci, rotavirus groups A and C, astrovirus, calicivirus, adenovirus types 40 and 41, enterovirus, hepatitis A virus and Cryptosporidum. Viruses were detected using an enzyme immunoassay for group A rotavirus, astrovirus and adenovirus 40 ℠41, and by RT-PCR for calicivirus, group C rotavirus, enterovirus and hepatitis A virus [4–6]. Campylobacter isolates were sent to the National Reference Centre for Campylobacter for identification and fingerprinting. Species identification was by means of phenotypic tests (API Campy; bioMe´rieux, Lyon, France), with use of PCR to differentiate Campylobacter jejuni from Camplylobacter coli [7]. Random amplified polymorphic DNA fingerprinting was performed with primer 3881, 5¢-AACGCGCAAC. Environmental investigation An environmental and microbiological investigation of the water distribution system was performed. The inspection concerned the water catchments, the water distribution and sewage system, agricultural practices and private households around the sources. Bacteriological (Escherichia coli, enterococci, faecal streptococci, thermotolerant coliforms, sulphite-reducing clostridia) and chemical (chlorine) analyses were performed on tap water and on the water sampled from the two groundwater sources. After 23 August, the water chlorine concentration and the presence of coliform bacteria were monitored daily at several sites of the water system before and after the resort treatment. Examinations for viruses (group A rotavirus, astrovirus, norovirus, enterovirus, hepatitis A virus) and Campylobacter spp. were undertaken on two 10-mL tap water samples collected inhouse on 23 August, and on two 10-mL samples from groundwater sources A and B on 25 August. Viruses were detected by RT-PCR, followed by hybridisation with a specific primer to improve sensitivity, while C. jejuni and C. coli were detected by PCR with specific primers. Group A rotavirus isolates detected in stools and in water were sequenced using Beg 9 and End 9 primers [8] and compared using the Infobiogene GCG software (http://www.infobiogen.fr). Bacteriological (revived 36C, revived 22C, thermotolerant coliforms, streptococci and faecal enterococci), physical (temperature, conductivity) and chemical (oxygen consumption, ammonium, nitrates, phosphorus) parameters from source A and the river were analysed. Trends were compared for an 8-week period between 21 December 2000 and 7 February 2001. RESULTS Epidemiology Between 1 August and 11 September 2000, 7104 Gourdon residents consulted a GP or the local hospital emergency department. Between 24 and 28 August, gastroenteritis accounted for 44% (479 ℠1097) of all medical visits, compared with 6% (126 ℠2185) during the first 15 days of August (Fig. 1). Among the 498 households contacted for the household survey, 198 were excluded (103 did not respond, 49 did not fulfil the inclusion criteria, 43 refused to participate, and three for unknown reasons). The questionnaire was completed for 300 households (709 individuals). Data were missing concerning gender for one individual and age for 11 individuals. Overall, 331 (46.7%; 95% CI 44.5–49.0) of respondents were male and 377 (53.2%; 95% CI 50.9–55.5) were female. The median age was 44 years (range < 1–94 years). The age and gender distribution of the sample was similar to that of the general population. During the outbreak period, 264 of the 709 respondents (attack rate (AR) 37.2%) had been ill, with 202 definite cases (AR 28.5%; 95% CI 24.6– 32.4) and 62 probable cases (AR 8.7%; 95% CI 6.5– 11.0). The AR was higher in females (31.8%) than in males (24.8%), and was highest in children aged < 6 years (42.1%, Table 1). The epidemic curve (Fig. 2) suggests that the outbreak started between 14 and 19 August. The shape of the epidemic curve in the retrospective cohort study was very similar to that based on the number of visits to GPs for gastroenteritis. Patients complained mainly of diarrhoea, abdominal pain and nausea (Table 2). No cases of bloody diarrhoea were notified. The mean duration of symptoms was 4.5 days (median 3, range 1–30 days). Among definite cases, 52% had visited a GP, 42% had taken sick leave for ‡ 1 day, 32% had required bed rest, and 2.9% (six patients) had been hospitalised. In total, 336 (47.4%) respondents had drunk tap water before 25 August, whereas 646 (91.1%) did not drink tap water after 25 August. The risk for illness, taking into account the sample cluster design effect of 1.4, was three-fold greater (95% CI 2.2–4.1) for those who had drunk tap water before 25 August, and increased with the amount of water consumed (chi-square trend: p < 0.0001; Table 3). The risk of gastroenteritis associated with drinking tap water did not change substantially after adjustment for age (RRMH 3.3, 95% CI 2.5–4.4), and there was no modification effect according to age. Food consumption was not associated with illness. Applying the AR of 37% to the total population of Gourdon exposed to the water supply (4888 residents + 2200 tourists), the number of individuals affected by this outbreak was estimated to be 2600 (95% CI 2400–2900), assuming an exposure to tap water and a risk for gastroenteritis among the tourists similar to that among the inhabitants. Microbiological stool analyses Among the 35 stool samples analysed for bacteria, 11 (31.5%) were positive for C. coli. Twenty-four of the 35 samples were also tested for viruses: rotavirus A was detected in 17 (71.0%) and norovirus in five (21.0%) samples, while nine (37.5%) samples were co-infected (four C. coli + rotavirus; one C. coli + norovirus; three rotavirus + norovirus; and one C. coli + rotavirus + norovirus). Of the 11 C. coli isolates, six were characterised as biotype 2, and were identical following RAPD analysis, while the remaining five were of biotype 1. Rotavirus typing by RTPCR identified a single genotype P-(8), G1. Molecular characterisation of the norovirus isolates identified two different patterns, with three isolates being identical and close to the Saragota strain (genogroup I) and two close to the new GGIIb variant (genogroup II). Environmental investigation All of the water samples from source B were negative, regardless of the site of the distribution system tested. Analyses of the water originating from source A, sampled after leaving the water-works on 23 and 25 August, revealed high concentrations of E. coli (2263 / 100 mL), enterococci (415 / 100 mL), total coliforms (90 / 100 mL), faecal streptococci (136 / 100 mL) and sulphite-reducing clostridia (10 / 100 mL), and an absence of chlorination. Rotavirus A genotype G1 was detected in the crude water sampled from source A. The analyses for enterovirus, astrovirus, norovirus, hepatitis A virus and Campylobacter were negative. Source A was situated in a valley containing a river, below a hamlet of 15 dwellings, including farms breeding cattle, pork and sheep. The source was not supposed to be supplied by the river. The river resulted from the convergence of about ten different streams, with one originating close to a slaughterhouse, and was exposed to agricultural run-off and manure effluent; sheep and cattle grazed in meadows bordering the river, and a large amount of sewage from water treatment facilities was sprayed on to land close to the river. Cowsheds were close to the water catchment area. The immediate protected area around source A was too small and littered with building rubbish. Air vents without protection in the door made the source accessible to animals and insects. Bacteriological analyses of the water samples from source A and the river revealed that the contamination was of both human and animal origin, with a higher concentration in the river. The levels and trends of the chemical parameters of source A were similar to those of the river, strongly suggesting that the river was connected to source A (Fig. 3). A retrospective review of routine data collected between 1991 and 2000 showed irregularities in the maintenance of the water distribution system and the water treatment plant, particularly during the month of August. In the hamlet, the sewage system of the 15 households did not comply with sanitary rules. All but three ejected waste either into a cesspool or superficially without treatment. Three regions of gene 9 of the rotavirus detected in stools and water were compared. Sequencing showed that the strains were closely related but not identical. Assuming a likely onset of the outbreak between 14 and 19 August, and taking into account a minimum incubation period of 2 days for rotavirus A, norovirus and C. coli, the contamination of the water distribution system probably occurred between 12 and 17 August. Contamination probably persisted until 28 August, as intensive chlorination, beginning on 25 August, was effective only from 28 August onwards (Fig. 2). DISCUSSION The present descriptive and retrospective cohort studies confirmed that the outbreak of infection in the community of Gourdon was related to the consumption of tap water and affected 2600 individuals. The outbreak involved multiple pathogens, with group A rotavirus and C. coli being the most frequent, followed by norovirus. Results of environmental investigations suggested that water contamination was related to the contamination of the source, combined with a failure of the water chlorination system. The association with tap water consumption, the dose-effect relationship, and the fact that multiple pathogens were involved, were all strongly in favour of tap water as the origin of infection. Moreover, the number of cases decreased dramatically after the water restriction advice and the implementation of control measures on the water supply. No food products explored in the study were associated with illness. An AR of 14.5% and 31% was observed among adults and children aged < 6 years, respectively, who reported that they had not consumed tap water. This relatively high rate could be explained by several phenomena. Many individuals may have consumed tap water without being aware of it, e.g., in drinks or ice cubes consumed outside of the home, via raw fruits and vegetables, dishes prepared with unboiled water, or while cleaning their teeth. It can be hypothesised that people who became ill without being aware of having consumed tap water were exposed to small amounts of water and had been subjected to a small infective dose. In addition, children, more than adults, may have been contaminated through person-to-person transmission (e.g., family contact), which is the most common route of transmission for rotavirus and calicivirus. Waterborne contamination can affect large numbers of individuals [3,9–11]. In the present study, assuming a 50% increase in the population during the tourist season, the number of affected people was estimated to be between 2400 and 2900. The cohort study showed that almost half (46.5%) of the cases consulted a GP. Applying this proportion to the estimated number of cases yielded an estimated 1116–1349 ill individuals who consulted a GP for gastroenteritis, similar to the number of consultations for gastroenteritis actually reported by the GPs (n = 1037). The stool sample analyses revealed the presence of multiple pathogens (rotavirus A, norovirus genogroups I and II and C. coli) as single or co-infections. Except for C. coli, these pathogens have been implicated in numerous waterborne outbreaks, suggesting an influx of sewage into the water supply system [10,12–18]. Rotaviruses are known to be responsible for large epidemics during winter seasons, affecting particularly children aged < 3 years. These epidemics are caused by the faecal route of contamination (personto-person transmission). The high proportion of stools positive for rotavirus (71%) is uncommon in waterborne outbreaks and probably reflects massive contamination [15,16,19]. Different molecular sequences of rotavirus A genotype G1 found in stools and water indicated a human faecal source of infection. In the present study, contamination of water by sewage could be linked with the large amount of sewage from the water purification plant sprayed on to land close to the river and the water catchments. In this outbreak, C. coli, which was identified in 32% of stools, was the second most frequent pathogen. To our knowledge, human C. coli contamination has not been described previously as a causative agent in waterborne outbreaks, unlike C. jejuni [11,20–25], although C. coli has been isolated from water [26]. The main symptoms observed (diarrhoea, abdominal pain, nausea) and the mean duration of the disease (4.5 days) were compatible with a Campylobacter infection. No bloody diarrhoea was reported, although bloody diarrhoea is not constant in Campylobacter infections [27]. The high proportion of individuals who had to take sick leave, and the fact that six individuals were hospitalised, suggested a relative severity of symptoms, which is much more compatible with Campylobacter infection than with virus infections. Among the hospitalised patients, 60% were infected with C. coli. Among patients whose stools were analysed and whose age was known, cases infected only with rotavirus were younger (mean age 49 years; range, 5 months to 95 years) than cases infected only with Campylobacter (mean age 66 years; range 23–88 years) or co-infected with rotavirus plus Campylobacter (mean age 69 years; range 2–91 years). Moreover, the proportion (52%) of medical visits was unusually high compared with the proportion of visits observed during other waterborne outbreaks [13,16,28]. Campylobacter needs specific atmospheric conditions for growth and survival and its detection can be very difficult [29,30]. Because of a rapid physiological transformation to a viable, though non-cultivable, state, it has been suggested that conventional culture methods often fail [31]. Nevertheless, in the present study, the failure to detect C. coli in water, and the detection of noroviruses in stools, could be caused by the small (10 mL) water samples examined. Ha¨nninen et al. [30] proposed that the volume used for detection of suspected pathogens in drinking water should be at least 8–10 L. Groundwater is considered widely to be microbiologically clean, and is usually used for drinking without treatment. However, groundwater can be contaminated by the environment [26,32]. Stanley et al. [33] showed that Campylobacter can occur in groundwater, with hydrological evidence suggesting that the source of contamination in their study was a dairy farm situated within the hydrological catchment area of the groundwater, which was confirmed when identical strains of C. jejuni were isolated from groundwater and the dairy herd. In the present study, Campylobacter was not identified from groundwater, and dairy herds and other domestic animals in farms around the source were not investigated. However, results of bacterial, chemical and physical water analyses suggested strongly that a river contaminated by agricultural run-off from surrounding sheep and cattle grazing meadows was contaminating the source. In addition, the immediate area of the source was insufficiently protected and vulnerable to the environment. Wild birds, rodents, pigs and cows are all hosts for C. coli and could contaminate surface water easily [26,27,34,35]. Campylobacters are sensitive to many environmental factors and to chlorine [33,36,37], but a failure of the chlorination system could have resulted in the presence of a high concentration of C. coli. Waterborne outbreaks are often caused by several factors that act together [18,20,23,38]. In some outbreaks, heavy rainfall has been responsible for increasing agricultural run-off into rivers, although this did not occur before the outbreak in Gourdon. This underlies the need for increased frequency of routine water quality monitoring, chlorination of the water supply according to environmental conditions, and implementation of preventive measures. This outbreak raised the question of the delay in raising the alert. An outbreak can be recognised by: (i) the poor quality of water detected during regular controls; (ii) complaints about water quality; (iii) an incident in the supply and ℠or sewage water system during maintenance work; and (iv) an increased number of cases of gastroenteritis. In most instances, waterborne outbreaks are suspected when there are already a large number of declared cases [3,12,18,28,38,39]. In the present outbreak, the initial warning was given by a GP who notified officials 5–10 days after the assumed date of onset. This highlights the role of GPs in the detection of an unusual event, particularly during the summer, when an increase in acute gastroenteritis is less expected than during winter [40,41]. In addition, a routine water quality monitoring system able to detect contamination of drinking water before the occurrence of illness and complaints about the water quality should be used as an early warning source. The retrospective cohort study of Gourdon residents allowed the scale and severity of the outbreak to be quantified precisely and the hypothesis that tap water was the vehicle of the outbreak to be tested. When provided with a cohort sample representative of the study population, the number of people affected can be estimated. However, as the exposure of the population was high, a relatively large sample size was needed to test the association between the disease and the exposure. Because of misclassification of exposure (unrecognised consumption of tap water), the attributable risk is probably underestimated. In the absence of obvious alternative explanations, water contamination is the most likely common cause of an explosive outbreak of AGE occurring in a large population. Tillett et al. [42] proposed categorising levels of evidence for waterborne outbreaks. A descriptive study suggesting a water-related outbreak, together with the identification of the same pathogen in stools and water, should be sufficient to conclude that an outbreak is associated strongly with water consumption. Faecal indicators of water quality are currently insufficiently sensitive for some pathogens, notably viruses and parasites (e.g., Cryptosporidium). Therefore, in the absence of evidence from an analytical study to demonstrate an association, the link between illness and water can only be classified as probable or possible. However, total coliform counts are good indicators of faecal contamination with other pathogens (e.g., rotaviruses, E. coli, Campylobacter) [43]. In conclusion, the waterborne outbreak described in the present study was most probably caused by faecal contamination of source A by groundwater, resulting from deficiencies in the maintenance of the water distribution system. Corrective measures were implemented based on the results of the investigation. To our knowledge, this is the first documented waterborne outbreak involving human C. coli infection. In-depth environmental investigations to identify the origin of water contamination are important in order to better understand microbial transmission from the environment and to implement appropriate controls and preventive measures. In addition, surveys on vehicles and vectors that transmit Campylobacter between hosts are crucial in order to better understand the epidemiology of these organisms.

In which street did this happen?

285

A large multi-pathogen waterborne community outbreak linked to faecal contamination of a groundwater system, France, 2000

A large waterborne outbreak of infection that occurred during August 2000 in a local community in France was investigated initially via a rapid survey of visits to local physicians. A retrospective cohort study was then conducted on a random cluster sample of residents. Of 709 residents interviewed, 202 (28.5%) were definite cases (at least three liquid stools ℠day or vomiting) and 62 (8.7%) were probable cases (less than three liquid stools ℠day or abdominal pain). Those who had drunk tap water had a threefold increased risk for illness (95% CI 2.4–4.0). The risk increased with the amount of water consumed (chi-square trend: p < 0.0001). Bacteriological analyses of stools were performed for 35 patients and virological analyses for 24 patients. Campylobacter coli, group A rotavirus and norovirus were detected in 31.5%, 71.0% and 21% of samples, respectively. An extensive environmental investigation concluded that a groundwater source to this community had probably been contaminated by agricultural run-off, and a failure in the chlorination system was identified. This is the first documented waterborne outbreak of infection involving human C. coli infections. A better understanding of the factors influencing campylobacter transmission between hosts is required. Keywords Campylobacter coli, epidemiology, gastroenteritis, norovirus, rotavirus, waterborne outbreak INTRODUCTION Contaminated drinking water causes extensive outbreaks of illness because of the large number of people served by water supply facilities. Detection of such outbreaks requires the identification of an increase in illness (usually the rate of gastrointestinal disease) in the exposed population and confirmation that water was the route of transmission [1]. Many countries have routine reporting systems for detecting foodborne and waterborne outbreaks of infection. However, in routine practice it is not easy to detect an increase in clinical cases and to link this increase to waterborne transmission; thus waterborne outbreaks are often unrecognised and underestimated [2]. Microbiological examination of water is more complex than examination of stools, and there is often a failure to detect pathogens in water. Norovirus, Cryptosporidium parvum, Giardia intestinalis and Campylobacter jejuni are the pathogens identified most frequently [3]. The present study describes a large waterborne community outbreak of infection with multiple pathogens, including Campylobacter coli. METHODS AND MATERIALS Background On 23 August 2000, a general practitioner (GP) informed the local district health department that 16 cases of acute gastroenteritis (AGE) had occurred among residents of a holiday camp in the Gourdon community since 21 August. Gourdon, which is located in the south-west of France, has a population of 4888 inhabitants, which increases by 50% in the July– August tourist season. Investigations at the holiday camp concluded that foodborne transmission was very unlikely, but found that tap water was highly contaminated with faecal coliforms. Preliminary information collected from physicians in Gourdon indicated an increase of the number of consultations for AGE. Gourdon tap water is supplied by two underground water facilities: facility A serving 3800 water consumers in Gourdon; and facility B serving 1200 water consumers in Gourdon and the surrounding communities. Water treatment in facility A consists of chlorination, which involves two automatic pumps working in rotation. The water then passes through two semi in-ground reservoir tanks (1100 m3 and 200 m3, respectively) before distribution. Facility B water is mixed with a fraction of facility A water in another 500 m3 semi in-ground reservoir tank before being distributed. This mixing with the treated water from facility A is intended to disinfect the water from facility B. Since the increase in AGE was noted only in Gourdon, contaminated tap water from facility A was suspected to be the origin of the outbreak. Therefore, the population of Gourdon was informed on 25 August to avoid the consumption of tap water or to boil it for 5 min. Survey of gastroenteritis-related medical consultations To confirm the outbreak, the 11 GPs in Gourdon and the local hospital emergency department were requested to provide the number of total daily visits and the number of cases of gastroenteritis cases occurring between 1 August and 11 September 2000. Retrospective cohort study A retrospective cohort study was performed on a sample of Gourdon households between 5 and 15 September 2000. Domestic households were selected randomly from the telephone directory. A definite case was defined as a Gourdon resident who had at least three liquid stools ℠day or vomiting between 1 August and 3 September, and who had been present in Gourdon for ‡ 1 day between 1 and 31 August; if less than three liquid stools ℠day or abdominal pain had occurred, the case was classified as probable. Exposure was defined as the consumption of tap water before 25 August 2000. A standardised telephone questionnaire was completed to obtain demographic data (gender, date of birth), food consumption (commercially prepared dishes, sea-food, raw milk, pastry), tap water consumption (before and after 25 August, daily number of glasses of water, consumption of boiled tap water), symptoms of gastroenteritis (date of onset, number of liquid stools ℠day, vomiting, nausea, abdominal pain, fever), medical care (in- or out-patient) and consequences on daily activities (sick leave, bed confinement) for each member of the selected household. Under the hypothesis that 60% of the population consumed tap water, an estimated 550 people were required for inclusion in the study to identify a relative risk (RR) of ‡ 3, with a 5% a risk and a power of 80%. Based on an average of two individuals per household (INSERM 1999 population census), 300 households were required for inclusion in the survey. The association between AGE and the consumption of tap water was assessed by calculating the risk ratio (RR) and 95% CI, taking into account a household cluster design effect. A dose-effect relationship was evaluated by analysing the RR trend for different categories of daily number of glasses of tap water consumed (chi-square trend). Proportions were compared using the corrected chi-square test, and means were compared using the Student test or Mann and Witney test. Data were analysed with Epi-Info 6.04c (CDC, Atlanta, GA, USA). The gender and age distribution of the cohort population were compared with that of the general population of Gourdon (INSERM 1999 population census) to assess whether it was representative. Under the hypothesis of representation, the total number of cases of gastroenteritis was estimated by applying the attack rate to the total population of Gourdon, including the tourist population (data provided by the Gourdon tourism office). The most likely period for contamination of the water distribution system was based on the epidemic curve and the minimal incubation period of the pathogens involved. Microbiological analyses of stools Stool samples were examined for the presence of Salmonella spp., Shigella spp., Staphylococcus aureus, Campylobacter spp., Yersinia enterocolitica, Escherichia coli, enterococci, rotavirus groups A and C, astrovirus, calicivirus, adenovirus types 40 and 41, enterovirus, hepatitis A virus and Cryptosporidum. Viruses were detected using an enzyme immunoassay for group A rotavirus, astrovirus and adenovirus 40 ℠41, and by RT-PCR for calicivirus, group C rotavirus, enterovirus and hepatitis A virus [4–6]. Campylobacter isolates were sent to the National Reference Centre for Campylobacter for identification and fingerprinting. Species identification was by means of phenotypic tests (API Campy; bioMe´rieux, Lyon, France), with use of PCR to differentiate Campylobacter jejuni from Camplylobacter coli [7]. Random amplified polymorphic DNA fingerprinting was performed with primer 3881, 5¢-AACGCGCAAC. Environmental investigation An environmental and microbiological investigation of the water distribution system was performed. The inspection concerned the water catchments, the water distribution and sewage system, agricultural practices and private households around the sources. Bacteriological (Escherichia coli, enterococci, faecal streptococci, thermotolerant coliforms, sulphite-reducing clostridia) and chemical (chlorine) analyses were performed on tap water and on the water sampled from the two groundwater sources. After 23 August, the water chlorine concentration and the presence of coliform bacteria were monitored daily at several sites of the water system before and after the resort treatment. Examinations for viruses (group A rotavirus, astrovirus, norovirus, enterovirus, hepatitis A virus) and Campylobacter spp. were undertaken on two 10-mL tap water samples collected inhouse on 23 August, and on two 10-mL samples from groundwater sources A and B on 25 August. Viruses were detected by RT-PCR, followed by hybridisation with a specific primer to improve sensitivity, while C. jejuni and C. coli were detected by PCR with specific primers. Group A rotavirus isolates detected in stools and in water were sequenced using Beg 9 and End 9 primers [8] and compared using the Infobiogene GCG software (http://www.infobiogen.fr). Bacteriological (revived 36C, revived 22C, thermotolerant coliforms, streptococci and faecal enterococci), physical (temperature, conductivity) and chemical (oxygen consumption, ammonium, nitrates, phosphorus) parameters from source A and the river were analysed. Trends were compared for an 8-week period between 21 December 2000 and 7 February 2001. RESULTS Epidemiology Between 1 August and 11 September 2000, 7104 Gourdon residents consulted a GP or the local hospital emergency department. Between 24 and 28 August, gastroenteritis accounted for 44% (479 ℠1097) of all medical visits, compared with 6% (126 ℠2185) during the first 15 days of August (Fig. 1). Among the 498 households contacted for the household survey, 198 were excluded (103 did not respond, 49 did not fulfil the inclusion criteria, 43 refused to participate, and three for unknown reasons). The questionnaire was completed for 300 households (709 individuals). Data were missing concerning gender for one individual and age for 11 individuals. Overall, 331 (46.7%; 95% CI 44.5–49.0) of respondents were male and 377 (53.2%; 95% CI 50.9–55.5) were female. The median age was 44 years (range < 1–94 years). The age and gender distribution of the sample was similar to that of the general population. During the outbreak period, 264 of the 709 respondents (attack rate (AR) 37.2%) had been ill, with 202 definite cases (AR 28.5%; 95% CI 24.6– 32.4) and 62 probable cases (AR 8.7%; 95% CI 6.5– 11.0). The AR was higher in females (31.8%) than in males (24.8%), and was highest in children aged < 6 years (42.1%, Table 1). The epidemic curve (Fig. 2) suggests that the outbreak started between 14 and 19 August. The shape of the epidemic curve in the retrospective cohort study was very similar to that based on the number of visits to GPs for gastroenteritis. Patients complained mainly of diarrhoea, abdominal pain and nausea (Table 2). No cases of bloody diarrhoea were notified. The mean duration of symptoms was 4.5 days (median 3, range 1–30 days). Among definite cases, 52% had visited a GP, 42% had taken sick leave for ‡ 1 day, 32% had required bed rest, and 2.9% (six patients) had been hospitalised. In total, 336 (47.4%) respondents had drunk tap water before 25 August, whereas 646 (91.1%) did not drink tap water after 25 August. The risk for illness, taking into account the sample cluster design effect of 1.4, was three-fold greater (95% CI 2.2–4.1) for those who had drunk tap water before 25 August, and increased with the amount of water consumed (chi-square trend: p < 0.0001; Table 3). The risk of gastroenteritis associated with drinking tap water did not change substantially after adjustment for age (RRMH 3.3, 95% CI 2.5–4.4), and there was no modification effect according to age. Food consumption was not associated with illness. Applying the AR of 37% to the total population of Gourdon exposed to the water supply (4888 residents + 2200 tourists), the number of individuals affected by this outbreak was estimated to be 2600 (95% CI 2400–2900), assuming an exposure to tap water and a risk for gastroenteritis among the tourists similar to that among the inhabitants. Microbiological stool analyses Among the 35 stool samples analysed for bacteria, 11 (31.5%) were positive for C. coli. Twenty-four of the 35 samples were also tested for viruses: rotavirus A was detected in 17 (71.0%) and norovirus in five (21.0%) samples, while nine (37.5%) samples were co-infected (four C. coli + rotavirus; one C. coli + norovirus; three rotavirus + norovirus; and one C. coli + rotavirus + norovirus). Of the 11 C. coli isolates, six were characterised as biotype 2, and were identical following RAPD analysis, while the remaining five were of biotype 1. Rotavirus typing by RTPCR identified a single genotype P-(8), G1. Molecular characterisation of the norovirus isolates identified two different patterns, with three isolates being identical and close to the Saragota strain (genogroup I) and two close to the new GGIIb variant (genogroup II). Environmental investigation All of the water samples from source B were negative, regardless of the site of the distribution system tested. Analyses of the water originating from source A, sampled after leaving the water-works on 23 and 25 August, revealed high concentrations of E. coli (2263 / 100 mL), enterococci (415 / 100 mL), total coliforms (90 / 100 mL), faecal streptococci (136 / 100 mL) and sulphite-reducing clostridia (10 / 100 mL), and an absence of chlorination. Rotavirus A genotype G1 was detected in the crude water sampled from source A. The analyses for enterovirus, astrovirus, norovirus, hepatitis A virus and Campylobacter were negative. Source A was situated in a valley containing a river, below a hamlet of 15 dwellings, including farms breeding cattle, pork and sheep. The source was not supposed to be supplied by the river. The river resulted from the convergence of about ten different streams, with one originating close to a slaughterhouse, and was exposed to agricultural run-off and manure effluent; sheep and cattle grazed in meadows bordering the river, and a large amount of sewage from water treatment facilities was sprayed on to land close to the river. Cowsheds were close to the water catchment area. The immediate protected area around source A was too small and littered with building rubbish. Air vents without protection in the door made the source accessible to animals and insects. Bacteriological analyses of the water samples from source A and the river revealed that the contamination was of both human and animal origin, with a higher concentration in the river. The levels and trends of the chemical parameters of source A were similar to those of the river, strongly suggesting that the river was connected to source A (Fig. 3). A retrospective review of routine data collected between 1991 and 2000 showed irregularities in the maintenance of the water distribution system and the water treatment plant, particularly during the month of August. In the hamlet, the sewage system of the 15 households did not comply with sanitary rules. All but three ejected waste either into a cesspool or superficially without treatment. Three regions of gene 9 of the rotavirus detected in stools and water were compared. Sequencing showed that the strains were closely related but not identical. Assuming a likely onset of the outbreak between 14 and 19 August, and taking into account a minimum incubation period of 2 days for rotavirus A, norovirus and C. coli, the contamination of the water distribution system probably occurred between 12 and 17 August. Contamination probably persisted until 28 August, as intensive chlorination, beginning on 25 August, was effective only from 28 August onwards (Fig. 2). DISCUSSION The present descriptive and retrospective cohort studies confirmed that the outbreak of infection in the community of Gourdon was related to the consumption of tap water and affected 2600 individuals. The outbreak involved multiple pathogens, with group A rotavirus and C. coli being the most frequent, followed by norovirus. Results of environmental investigations suggested that water contamination was related to the contamination of the source, combined with a failure of the water chlorination system. The association with tap water consumption, the dose-effect relationship, and the fact that multiple pathogens were involved, were all strongly in favour of tap water as the origin of infection. Moreover, the number of cases decreased dramatically after the water restriction advice and the implementation of control measures on the water supply. No food products explored in the study were associated with illness. An AR of 14.5% and 31% was observed among adults and children aged < 6 years, respectively, who reported that they had not consumed tap water. This relatively high rate could be explained by several phenomena. Many individuals may have consumed tap water without being aware of it, e.g., in drinks or ice cubes consumed outside of the home, via raw fruits and vegetables, dishes prepared with unboiled water, or while cleaning their teeth. It can be hypothesised that people who became ill without being aware of having consumed tap water were exposed to small amounts of water and had been subjected to a small infective dose. In addition, children, more than adults, may have been contaminated through person-to-person transmission (e.g., family contact), which is the most common route of transmission for rotavirus and calicivirus. Waterborne contamination can affect large numbers of individuals [3,9–11]. In the present study, assuming a 50% increase in the population during the tourist season, the number of affected people was estimated to be between 2400 and 2900. The cohort study showed that almost half (46.5%) of the cases consulted a GP. Applying this proportion to the estimated number of cases yielded an estimated 1116–1349 ill individuals who consulted a GP for gastroenteritis, similar to the number of consultations for gastroenteritis actually reported by the GPs (n = 1037). The stool sample analyses revealed the presence of multiple pathogens (rotavirus A, norovirus genogroups I and II and C. coli) as single or co-infections. Except for C. coli, these pathogens have been implicated in numerous waterborne outbreaks, suggesting an influx of sewage into the water supply system [10,12–18]. Rotaviruses are known to be responsible for large epidemics during winter seasons, affecting particularly children aged < 3 years. These epidemics are caused by the faecal route of contamination (personto-person transmission). The high proportion of stools positive for rotavirus (71%) is uncommon in waterborne outbreaks and probably reflects massive contamination [15,16,19]. Different molecular sequences of rotavirus A genotype G1 found in stools and water indicated a human faecal source of infection. In the present study, contamination of water by sewage could be linked with the large amount of sewage from the water purification plant sprayed on to land close to the river and the water catchments. In this outbreak, C. coli, which was identified in 32% of stools, was the second most frequent pathogen. To our knowledge, human C. coli contamination has not been described previously as a causative agent in waterborne outbreaks, unlike C. jejuni [11,20–25], although C. coli has been isolated from water [26]. The main symptoms observed (diarrhoea, abdominal pain, nausea) and the mean duration of the disease (4.5 days) were compatible with a Campylobacter infection. No bloody diarrhoea was reported, although bloody diarrhoea is not constant in Campylobacter infections [27]. The high proportion of individuals who had to take sick leave, and the fact that six individuals were hospitalised, suggested a relative severity of symptoms, which is much more compatible with Campylobacter infection than with virus infections. Among the hospitalised patients, 60% were infected with C. coli. Among patients whose stools were analysed and whose age was known, cases infected only with rotavirus were younger (mean age 49 years; range, 5 months to 95 years) than cases infected only with Campylobacter (mean age 66 years; range 23–88 years) or co-infected with rotavirus plus Campylobacter (mean age 69 years; range 2–91 years). Moreover, the proportion (52%) of medical visits was unusually high compared with the proportion of visits observed during other waterborne outbreaks [13,16,28]. Campylobacter needs specific atmospheric conditions for growth and survival and its detection can be very difficult [29,30]. Because of a rapid physiological transformation to a viable, though non-cultivable, state, it has been suggested that conventional culture methods often fail [31]. Nevertheless, in the present study, the failure to detect C. coli in water, and the detection of noroviruses in stools, could be caused by the small (10 mL) water samples examined. Ha¨nninen et al. [30] proposed that the volume used for detection of suspected pathogens in drinking water should be at least 8–10 L. Groundwater is considered widely to be microbiologically clean, and is usually used for drinking without treatment. However, groundwater can be contaminated by the environment [26,32]. Stanley et al. [33] showed that Campylobacter can occur in groundwater, with hydrological evidence suggesting that the source of contamination in their study was a dairy farm situated within the hydrological catchment area of the groundwater, which was confirmed when identical strains of C. jejuni were isolated from groundwater and the dairy herd. In the present study, Campylobacter was not identified from groundwater, and dairy herds and other domestic animals in farms around the source were not investigated. However, results of bacterial, chemical and physical water analyses suggested strongly that a river contaminated by agricultural run-off from surrounding sheep and cattle grazing meadows was contaminating the source. In addition, the immediate area of the source was insufficiently protected and vulnerable to the environment. Wild birds, rodents, pigs and cows are all hosts for C. coli and could contaminate surface water easily [26,27,34,35]. Campylobacters are sensitive to many environmental factors and to chlorine [33,36,37], but a failure of the chlorination system could have resulted in the presence of a high concentration of C. coli. Waterborne outbreaks are often caused by several factors that act together [18,20,23,38]. In some outbreaks, heavy rainfall has been responsible for increasing agricultural run-off into rivers, although this did not occur before the outbreak in Gourdon. This underlies the need for increased frequency of routine water quality monitoring, chlorination of the water supply according to environmental conditions, and implementation of preventive measures. This outbreak raised the question of the delay in raising the alert. An outbreak can be recognised by: (i) the poor quality of water detected during regular controls; (ii) complaints about water quality; (iii) an incident in the supply and ℠or sewage water system during maintenance work; and (iv) an increased number of cases of gastroenteritis. In most instances, waterborne outbreaks are suspected when there are already a large number of declared cases [3,12,18,28,38,39]. In the present outbreak, the initial warning was given by a GP who notified officials 5–10 days after the assumed date of onset. This highlights the role of GPs in the detection of an unusual event, particularly during the summer, when an increase in acute gastroenteritis is less expected than during winter [40,41]. In addition, a routine water quality monitoring system able to detect contamination of drinking water before the occurrence of illness and complaints about the water quality should be used as an early warning source. The retrospective cohort study of Gourdon residents allowed the scale and severity of the outbreak to be quantified precisely and the hypothesis that tap water was the vehicle of the outbreak to be tested. When provided with a cohort sample representative of the study population, the number of people affected can be estimated. However, as the exposure of the population was high, a relatively large sample size was needed to test the association between the disease and the exposure. Because of misclassification of exposure (unrecognised consumption of tap water), the attributable risk is probably underestimated. In the absence of obvious alternative explanations, water contamination is the most likely common cause of an explosive outbreak of AGE occurring in a large population. Tillett et al. [42] proposed categorising levels of evidence for waterborne outbreaks. A descriptive study suggesting a water-related outbreak, together with the identification of the same pathogen in stools and water, should be sufficient to conclude that an outbreak is associated strongly with water consumption. Faecal indicators of water quality are currently insufficiently sensitive for some pathogens, notably viruses and parasites (e.g., Cryptosporidium). Therefore, in the absence of evidence from an analytical study to demonstrate an association, the link between illness and water can only be classified as probable or possible. However, total coliform counts are good indicators of faecal contamination with other pathogens (e.g., rotaviruses, E. coli, Campylobacter) [43]. In conclusion, the waterborne outbreak described in the present study was most probably caused by faecal contamination of source A by groundwater, resulting from deficiencies in the maintenance of the water distribution system. Corrective measures were implemented based on the results of the investigation. To our knowledge, this is the first documented waterborne outbreak involving human C. coli infection. In-depth environmental investigations to identify the origin of water contamination are important in order to better understand microbial transmission from the environment and to implement appropriate controls and preventive measures. In addition, surveys on vehicles and vectors that transmit Campylobacter between hosts are crucial in order to better understand the epidemiology of these organisms.

In which city did this happen?

286

A large multi-pathogen waterborne community outbreak linked to faecal contamination of a groundwater system, France, 2000

A large waterborne outbreak of infection that occurred during August 2000 in a local community in France was investigated initially via a rapid survey of visits to local physicians. A retrospective cohort study was then conducted on a random cluster sample of residents. Of 709 residents interviewed, 202 (28.5%) were definite cases (at least three liquid stools ℠day or vomiting) and 62 (8.7%) were probable cases (less than three liquid stools ℠day or abdominal pain). Those who had drunk tap water had a threefold increased risk for illness (95% CI 2.4–4.0). The risk increased with the amount of water consumed (chi-square trend: p < 0.0001). Bacteriological analyses of stools were performed for 35 patients and virological analyses for 24 patients. Campylobacter coli, group A rotavirus and norovirus were detected in 31.5%, 71.0% and 21% of samples, respectively. An extensive environmental investigation concluded that a groundwater source to this community had probably been contaminated by agricultural run-off, and a failure in the chlorination system was identified. This is the first documented waterborne outbreak of infection involving human C. coli infections. A better understanding of the factors influencing campylobacter transmission between hosts is required. Keywords Campylobacter coli, epidemiology, gastroenteritis, norovirus, rotavirus, waterborne outbreak INTRODUCTION Contaminated drinking water causes extensive outbreaks of illness because of the large number of people served by water supply facilities. Detection of such outbreaks requires the identification of an increase in illness (usually the rate of gastrointestinal disease) in the exposed population and confirmation that water was the route of transmission [1]. Many countries have routine reporting systems for detecting foodborne and waterborne outbreaks of infection. However, in routine practice it is not easy to detect an increase in clinical cases and to link this increase to waterborne transmission; thus waterborne outbreaks are often unrecognised and underestimated [2]. Microbiological examination of water is more complex than examination of stools, and there is often a failure to detect pathogens in water. Norovirus, Cryptosporidium parvum, Giardia intestinalis and Campylobacter jejuni are the pathogens identified most frequently [3]. The present study describes a large waterborne community outbreak of infection with multiple pathogens, including Campylobacter coli. METHODS AND MATERIALS Background On 23 August 2000, a general practitioner (GP) informed the local district health department that 16 cases of acute gastroenteritis (AGE) had occurred among residents of a holiday camp in the Gourdon community since 21 August. Gourdon, which is located in the south-west of France, has a population of 4888 inhabitants, which increases by 50% in the July– August tourist season. Investigations at the holiday camp concluded that foodborne transmission was very unlikely, but found that tap water was highly contaminated with faecal coliforms. Preliminary information collected from physicians in Gourdon indicated an increase of the number of consultations for AGE. Gourdon tap water is supplied by two underground water facilities: facility A serving 3800 water consumers in Gourdon; and facility B serving 1200 water consumers in Gourdon and the surrounding communities. Water treatment in facility A consists of chlorination, which involves two automatic pumps working in rotation. The water then passes through two semi in-ground reservoir tanks (1100 m3 and 200 m3, respectively) before distribution. Facility B water is mixed with a fraction of facility A water in another 500 m3 semi in-ground reservoir tank before being distributed. This mixing with the treated water from facility A is intended to disinfect the water from facility B. Since the increase in AGE was noted only in Gourdon, contaminated tap water from facility A was suspected to be the origin of the outbreak. Therefore, the population of Gourdon was informed on 25 August to avoid the consumption of tap water or to boil it for 5 min. Survey of gastroenteritis-related medical consultations To confirm the outbreak, the 11 GPs in Gourdon and the local hospital emergency department were requested to provide the number of total daily visits and the number of cases of gastroenteritis cases occurring between 1 August and 11 September 2000. Retrospective cohort study A retrospective cohort study was performed on a sample of Gourdon households between 5 and 15 September 2000. Domestic households were selected randomly from the telephone directory. A definite case was defined as a Gourdon resident who had at least three liquid stools ℠day or vomiting between 1 August and 3 September, and who had been present in Gourdon for ‡ 1 day between 1 and 31 August; if less than three liquid stools ℠day or abdominal pain had occurred, the case was classified as probable. Exposure was defined as the consumption of tap water before 25 August 2000. A standardised telephone questionnaire was completed to obtain demographic data (gender, date of birth), food consumption (commercially prepared dishes, sea-food, raw milk, pastry), tap water consumption (before and after 25 August, daily number of glasses of water, consumption of boiled tap water), symptoms of gastroenteritis (date of onset, number of liquid stools ℠day, vomiting, nausea, abdominal pain, fever), medical care (in- or out-patient) and consequences on daily activities (sick leave, bed confinement) for each member of the selected household. Under the hypothesis that 60% of the population consumed tap water, an estimated 550 people were required for inclusion in the study to identify a relative risk (RR) of ‡ 3, with a 5% a risk and a power of 80%. Based on an average of two individuals per household (INSERM 1999 population census), 300 households were required for inclusion in the survey. The association between AGE and the consumption of tap water was assessed by calculating the risk ratio (RR) and 95% CI, taking into account a household cluster design effect. A dose-effect relationship was evaluated by analysing the RR trend for different categories of daily number of glasses of tap water consumed (chi-square trend). Proportions were compared using the corrected chi-square test, and means were compared using the Student test or Mann and Witney test. Data were analysed with Epi-Info 6.04c (CDC, Atlanta, GA, USA). The gender and age distribution of the cohort population were compared with that of the general population of Gourdon (INSERM 1999 population census) to assess whether it was representative. Under the hypothesis of representation, the total number of cases of gastroenteritis was estimated by applying the attack rate to the total population of Gourdon, including the tourist population (data provided by the Gourdon tourism office). The most likely period for contamination of the water distribution system was based on the epidemic curve and the minimal incubation period of the pathogens involved. Microbiological analyses of stools Stool samples were examined for the presence of Salmonella spp., Shigella spp., Staphylococcus aureus, Campylobacter spp., Yersinia enterocolitica, Escherichia coli, enterococci, rotavirus groups A and C, astrovirus, calicivirus, adenovirus types 40 and 41, enterovirus, hepatitis A virus and Cryptosporidum. Viruses were detected using an enzyme immunoassay for group A rotavirus, astrovirus and adenovirus 40 ℠41, and by RT-PCR for calicivirus, group C rotavirus, enterovirus and hepatitis A virus [4–6]. Campylobacter isolates were sent to the National Reference Centre for Campylobacter for identification and fingerprinting. Species identification was by means of phenotypic tests (API Campy; bioMe´rieux, Lyon, France), with use of PCR to differentiate Campylobacter jejuni from Camplylobacter coli [7]. Random amplified polymorphic DNA fingerprinting was performed with primer 3881, 5¢-AACGCGCAAC. Environmental investigation An environmental and microbiological investigation of the water distribution system was performed. The inspection concerned the water catchments, the water distribution and sewage system, agricultural practices and private households around the sources. Bacteriological (Escherichia coli, enterococci, faecal streptococci, thermotolerant coliforms, sulphite-reducing clostridia) and chemical (chlorine) analyses were performed on tap water and on the water sampled from the two groundwater sources. After 23 August, the water chlorine concentration and the presence of coliform bacteria were monitored daily at several sites of the water system before and after the resort treatment. Examinations for viruses (group A rotavirus, astrovirus, norovirus, enterovirus, hepatitis A virus) and Campylobacter spp. were undertaken on two 10-mL tap water samples collected inhouse on 23 August, and on two 10-mL samples from groundwater sources A and B on 25 August. Viruses were detected by RT-PCR, followed by hybridisation with a specific primer to improve sensitivity, while C. jejuni and C. coli were detected by PCR with specific primers. Group A rotavirus isolates detected in stools and in water were sequenced using Beg 9 and End 9 primers [8] and compared using the Infobiogene GCG software (http://www.infobiogen.fr). Bacteriological (revived 36C, revived 22C, thermotolerant coliforms, streptococci and faecal enterococci), physical (temperature, conductivity) and chemical (oxygen consumption, ammonium, nitrates, phosphorus) parameters from source A and the river were analysed. Trends were compared for an 8-week period between 21 December 2000 and 7 February 2001. RESULTS Epidemiology Between 1 August and 11 September 2000, 7104 Gourdon residents consulted a GP or the local hospital emergency department. Between 24 and 28 August, gastroenteritis accounted for 44% (479 ℠1097) of all medical visits, compared with 6% (126 ℠2185) during the first 15 days of August (Fig. 1). Among the 498 households contacted for the household survey, 198 were excluded (103 did not respond, 49 did not fulfil the inclusion criteria, 43 refused to participate, and three for unknown reasons). The questionnaire was completed for 300 households (709 individuals). Data were missing concerning gender for one individual and age for 11 individuals. Overall, 331 (46.7%; 95% CI 44.5–49.0) of respondents were male and 377 (53.2%; 95% CI 50.9–55.5) were female. The median age was 44 years (range < 1–94 years). The age and gender distribution of the sample was similar to that of the general population. During the outbreak period, 264 of the 709 respondents (attack rate (AR) 37.2%) had been ill, with 202 definite cases (AR 28.5%; 95% CI 24.6– 32.4) and 62 probable cases (AR 8.7%; 95% CI 6.5– 11.0). The AR was higher in females (31.8%) than in males (24.8%), and was highest in children aged < 6 years (42.1%, Table 1). The epidemic curve (Fig. 2) suggests that the outbreak started between 14 and 19 August. The shape of the epidemic curve in the retrospective cohort study was very similar to that based on the number of visits to GPs for gastroenteritis. Patients complained mainly of diarrhoea, abdominal pain and nausea (Table 2). No cases of bloody diarrhoea were notified. The mean duration of symptoms was 4.5 days (median 3, range 1–30 days). Among definite cases, 52% had visited a GP, 42% had taken sick leave for ‡ 1 day, 32% had required bed rest, and 2.9% (six patients) had been hospitalised. In total, 336 (47.4%) respondents had drunk tap water before 25 August, whereas 646 (91.1%) did not drink tap water after 25 August. The risk for illness, taking into account the sample cluster design effect of 1.4, was three-fold greater (95% CI 2.2–4.1) for those who had drunk tap water before 25 August, and increased with the amount of water consumed (chi-square trend: p < 0.0001; Table 3). The risk of gastroenteritis associated with drinking tap water did not change substantially after adjustment for age (RRMH 3.3, 95% CI 2.5–4.4), and there was no modification effect according to age. Food consumption was not associated with illness. Applying the AR of 37% to the total population of Gourdon exposed to the water supply (4888 residents + 2200 tourists), the number of individuals affected by this outbreak was estimated to be 2600 (95% CI 2400–2900), assuming an exposure to tap water and a risk for gastroenteritis among the tourists similar to that among the inhabitants. Microbiological stool analyses Among the 35 stool samples analysed for bacteria, 11 (31.5%) were positive for C. coli. Twenty-four of the 35 samples were also tested for viruses: rotavirus A was detected in 17 (71.0%) and norovirus in five (21.0%) samples, while nine (37.5%) samples were co-infected (four C. coli + rotavirus; one C. coli + norovirus; three rotavirus + norovirus; and one C. coli + rotavirus + norovirus). Of the 11 C. coli isolates, six were characterised as biotype 2, and were identical following RAPD analysis, while the remaining five were of biotype 1. Rotavirus typing by RTPCR identified a single genotype P-(8), G1. Molecular characterisation of the norovirus isolates identified two different patterns, with three isolates being identical and close to the Saragota strain (genogroup I) and two close to the new GGIIb variant (genogroup II). Environmental investigation All of the water samples from source B were negative, regardless of the site of the distribution system tested. Analyses of the water originating from source A, sampled after leaving the water-works on 23 and 25 August, revealed high concentrations of E. coli (2263 / 100 mL), enterococci (415 / 100 mL), total coliforms (90 / 100 mL), faecal streptococci (136 / 100 mL) and sulphite-reducing clostridia (10 / 100 mL), and an absence of chlorination. Rotavirus A genotype G1 was detected in the crude water sampled from source A. The analyses for enterovirus, astrovirus, norovirus, hepatitis A virus and Campylobacter were negative. Source A was situated in a valley containing a river, below a hamlet of 15 dwellings, including farms breeding cattle, pork and sheep. The source was not supposed to be supplied by the river. The river resulted from the convergence of about ten different streams, with one originating close to a slaughterhouse, and was exposed to agricultural run-off and manure effluent; sheep and cattle grazed in meadows bordering the river, and a large amount of sewage from water treatment facilities was sprayed on to land close to the river. Cowsheds were close to the water catchment area. The immediate protected area around source A was too small and littered with building rubbish. Air vents without protection in the door made the source accessible to animals and insects. Bacteriological analyses of the water samples from source A and the river revealed that the contamination was of both human and animal origin, with a higher concentration in the river. The levels and trends of the chemical parameters of source A were similar to those of the river, strongly suggesting that the river was connected to source A (Fig. 3). A retrospective review of routine data collected between 1991 and 2000 showed irregularities in the maintenance of the water distribution system and the water treatment plant, particularly during the month of August. In the hamlet, the sewage system of the 15 households did not comply with sanitary rules. All but three ejected waste either into a cesspool or superficially without treatment. Three regions of gene 9 of the rotavirus detected in stools and water were compared. Sequencing showed that the strains were closely related but not identical. Assuming a likely onset of the outbreak between 14 and 19 August, and taking into account a minimum incubation period of 2 days for rotavirus A, norovirus and C. coli, the contamination of the water distribution system probably occurred between 12 and 17 August. Contamination probably persisted until 28 August, as intensive chlorination, beginning on 25 August, was effective only from 28 August onwards (Fig. 2). DISCUSSION The present descriptive and retrospective cohort studies confirmed that the outbreak of infection in the community of Gourdon was related to the consumption of tap water and affected 2600 individuals. The outbreak involved multiple pathogens, with group A rotavirus and C. coli being the most frequent, followed by norovirus. Results of environmental investigations suggested that water contamination was related to the contamination of the source, combined with a failure of the water chlorination system. The association with tap water consumption, the dose-effect relationship, and the fact that multiple pathogens were involved, were all strongly in favour of tap water as the origin of infection. Moreover, the number of cases decreased dramatically after the water restriction advice and the implementation of control measures on the water supply. No food products explored in the study were associated with illness. An AR of 14.5% and 31% was observed among adults and children aged < 6 years, respectively, who reported that they had not consumed tap water. This relatively high rate could be explained by several phenomena. Many individuals may have consumed tap water without being aware of it, e.g., in drinks or ice cubes consumed outside of the home, via raw fruits and vegetables, dishes prepared with unboiled water, or while cleaning their teeth. It can be hypothesised that people who became ill without being aware of having consumed tap water were exposed to small amounts of water and had been subjected to a small infective dose. In addition, children, more than adults, may have been contaminated through person-to-person transmission (e.g., family contact), which is the most common route of transmission for rotavirus and calicivirus. Waterborne contamination can affect large numbers of individuals [3,9–11]. In the present study, assuming a 50% increase in the population during the tourist season, the number of affected people was estimated to be between 2400 and 2900. The cohort study showed that almost half (46.5%) of the cases consulted a GP. Applying this proportion to the estimated number of cases yielded an estimated 1116–1349 ill individuals who consulted a GP for gastroenteritis, similar to the number of consultations for gastroenteritis actually reported by the GPs (n = 1037). The stool sample analyses revealed the presence of multiple pathogens (rotavirus A, norovirus genogroups I and II and C. coli) as single or co-infections. Except for C. coli, these pathogens have been implicated in numerous waterborne outbreaks, suggesting an influx of sewage into the water supply system [10,12–18]. Rotaviruses are known to be responsible for large epidemics during winter seasons, affecting particularly children aged < 3 years. These epidemics are caused by the faecal route of contamination (personto-person transmission). The high proportion of stools positive for rotavirus (71%) is uncommon in waterborne outbreaks and probably reflects massive contamination [15,16,19]. Different molecular sequences of rotavirus A genotype G1 found in stools and water indicated a human faecal source of infection. In the present study, contamination of water by sewage could be linked with the large amount of sewage from the water purification plant sprayed on to land close to the river and the water catchments. In this outbreak, C. coli, which was identified in 32% of stools, was the second most frequent pathogen. To our knowledge, human C. coli contamination has not been described previously as a causative agent in waterborne outbreaks, unlike C. jejuni [11,20–25], although C. coli has been isolated from water [26]. The main symptoms observed (diarrhoea, abdominal pain, nausea) and the mean duration of the disease (4.5 days) were compatible with a Campylobacter infection. No bloody diarrhoea was reported, although bloody diarrhoea is not constant in Campylobacter infections [27]. The high proportion of individuals who had to take sick leave, and the fact that six individuals were hospitalised, suggested a relative severity of symptoms, which is much more compatible with Campylobacter infection than with virus infections. Among the hospitalised patients, 60% were infected with C. coli. Among patients whose stools were analysed and whose age was known, cases infected only with rotavirus were younger (mean age 49 years; range, 5 months to 95 years) than cases infected only with Campylobacter (mean age 66 years; range 23–88 years) or co-infected with rotavirus plus Campylobacter (mean age 69 years; range 2–91 years). Moreover, the proportion (52%) of medical visits was unusually high compared with the proportion of visits observed during other waterborne outbreaks [13,16,28]. Campylobacter needs specific atmospheric conditions for growth and survival and its detection can be very difficult [29,30]. Because of a rapid physiological transformation to a viable, though non-cultivable, state, it has been suggested that conventional culture methods often fail [31]. Nevertheless, in the present study, the failure to detect C. coli in water, and the detection of noroviruses in stools, could be caused by the small (10 mL) water samples examined. Ha¨nninen et al. [30] proposed that the volume used for detection of suspected pathogens in drinking water should be at least 8–10 L. Groundwater is considered widely to be microbiologically clean, and is usually used for drinking without treatment. However, groundwater can be contaminated by the environment [26,32]. Stanley et al. [33] showed that Campylobacter can occur in groundwater, with hydrological evidence suggesting that the source of contamination in their study was a dairy farm situated within the hydrological catchment area of the groundwater, which was confirmed when identical strains of C. jejuni were isolated from groundwater and the dairy herd. In the present study, Campylobacter was not identified from groundwater, and dairy herds and other domestic animals in farms around the source were not investigated. However, results of bacterial, chemical and physical water analyses suggested strongly that a river contaminated by agricultural run-off from surrounding sheep and cattle grazing meadows was contaminating the source. In addition, the immediate area of the source was insufficiently protected and vulnerable to the environment. Wild birds, rodents, pigs and cows are all hosts for C. coli and could contaminate surface water easily [26,27,34,35]. Campylobacters are sensitive to many environmental factors and to chlorine [33,36,37], but a failure of the chlorination system could have resulted in the presence of a high concentration of C. coli. Waterborne outbreaks are often caused by several factors that act together [18,20,23,38]. In some outbreaks, heavy rainfall has been responsible for increasing agricultural run-off into rivers, although this did not occur before the outbreak in Gourdon. This underlies the need for increased frequency of routine water quality monitoring, chlorination of the water supply according to environmental conditions, and implementation of preventive measures. This outbreak raised the question of the delay in raising the alert. An outbreak can be recognised by: (i) the poor quality of water detected during regular controls; (ii) complaints about water quality; (iii) an incident in the supply and ℠or sewage water system during maintenance work; and (iv) an increased number of cases of gastroenteritis. In most instances, waterborne outbreaks are suspected when there are already a large number of declared cases [3,12,18,28,38,39]. In the present outbreak, the initial warning was given by a GP who notified officials 5–10 days after the assumed date of onset. This highlights the role of GPs in the detection of an unusual event, particularly during the summer, when an increase in acute gastroenteritis is less expected than during winter [40,41]. In addition, a routine water quality monitoring system able to detect contamination of drinking water before the occurrence of illness and complaints about the water quality should be used as an early warning source. The retrospective cohort study of Gourdon residents allowed the scale and severity of the outbreak to be quantified precisely and the hypothesis that tap water was the vehicle of the outbreak to be tested. When provided with a cohort sample representative of the study population, the number of people affected can be estimated. However, as the exposure of the population was high, a relatively large sample size was needed to test the association between the disease and the exposure. Because of misclassification of exposure (unrecognised consumption of tap water), the attributable risk is probably underestimated. In the absence of obvious alternative explanations, water contamination is the most likely common cause of an explosive outbreak of AGE occurring in a large population. Tillett et al. [42] proposed categorising levels of evidence for waterborne outbreaks. A descriptive study suggesting a water-related outbreak, together with the identification of the same pathogen in stools and water, should be sufficient to conclude that an outbreak is associated strongly with water consumption. Faecal indicators of water quality are currently insufficiently sensitive for some pathogens, notably viruses and parasites (e.g., Cryptosporidium). Therefore, in the absence of evidence from an analytical study to demonstrate an association, the link between illness and water can only be classified as probable or possible. However, total coliform counts are good indicators of faecal contamination with other pathogens (e.g., rotaviruses, E. coli, Campylobacter) [43]. In conclusion, the waterborne outbreak described in the present study was most probably caused by faecal contamination of source A by groundwater, resulting from deficiencies in the maintenance of the water distribution system. Corrective measures were implemented based on the results of the investigation. To our knowledge, this is the first documented waterborne outbreak involving human C. coli infection. In-depth environmental investigations to identify the origin of water contamination are important in order to better understand microbial transmission from the environment and to implement appropriate controls and preventive measures. In addition, surveys on vehicles and vectors that transmit Campylobacter between hosts are crucial in order to better understand the epidemiology of these organisms.

In which region did this happen?

287

A large multi-pathogen waterborne community outbreak linked to faecal contamination of a groundwater system, France, 2000

A large waterborne outbreak of infection that occurred during August 2000 in a local community in France was investigated initially via a rapid survey of visits to local physicians. A retrospective cohort study was then conducted on a random cluster sample of residents. Of 709 residents interviewed, 202 (28.5%) were definite cases (at least three liquid stools ℠day or vomiting) and 62 (8.7%) were probable cases (less than three liquid stools ℠day or abdominal pain). Those who had drunk tap water had a threefold increased risk for illness (95% CI 2.4–4.0). The risk increased with the amount of water consumed (chi-square trend: p < 0.0001). Bacteriological analyses of stools were performed for 35 patients and virological analyses for 24 patients. Campylobacter coli, group A rotavirus and norovirus were detected in 31.5%, 71.0% and 21% of samples, respectively. An extensive environmental investigation concluded that a groundwater source to this community had probably been contaminated by agricultural run-off, and a failure in the chlorination system was identified. This is the first documented waterborne outbreak of infection involving human C. coli infections. A better understanding of the factors influencing campylobacter transmission between hosts is required. Keywords Campylobacter coli, epidemiology, gastroenteritis, norovirus, rotavirus, waterborne outbreak INTRODUCTION Contaminated drinking water causes extensive outbreaks of illness because of the large number of people served by water supply facilities. Detection of such outbreaks requires the identification of an increase in illness (usually the rate of gastrointestinal disease) in the exposed population and confirmation that water was the route of transmission [1]. Many countries have routine reporting systems for detecting foodborne and waterborne outbreaks of infection. However, in routine practice it is not easy to detect an increase in clinical cases and to link this increase to waterborne transmission; thus waterborne outbreaks are often unrecognised and underestimated [2]. Microbiological examination of water is more complex than examination of stools, and there is often a failure to detect pathogens in water. Norovirus, Cryptosporidium parvum, Giardia intestinalis and Campylobacter jejuni are the pathogens identified most frequently [3]. The present study describes a large waterborne community outbreak of infection with multiple pathogens, including Campylobacter coli. METHODS AND MATERIALS Background On 23 August 2000, a general practitioner (GP) informed the local district health department that 16 cases of acute gastroenteritis (AGE) had occurred among residents of a holiday camp in the Gourdon community since 21 August. Gourdon, which is located in the south-west of France, has a population of 4888 inhabitants, which increases by 50% in the July– August tourist season. Investigations at the holiday camp concluded that foodborne transmission was very unlikely, but found that tap water was highly contaminated with faecal coliforms. Preliminary information collected from physicians in Gourdon indicated an increase of the number of consultations for AGE. Gourdon tap water is supplied by two underground water facilities: facility A serving 3800 water consumers in Gourdon; and facility B serving 1200 water consumers in Gourdon and the surrounding communities. Water treatment in facility A consists of chlorination, which involves two automatic pumps working in rotation. The water then passes through two semi in-ground reservoir tanks (1100 m3 and 200 m3, respectively) before distribution. Facility B water is mixed with a fraction of facility A water in another 500 m3 semi in-ground reservoir tank before being distributed. This mixing with the treated water from facility A is intended to disinfect the water from facility B. Since the increase in AGE was noted only in Gourdon, contaminated tap water from facility A was suspected to be the origin of the outbreak. Therefore, the population of Gourdon was informed on 25 August to avoid the consumption of tap water or to boil it for 5 min. Survey of gastroenteritis-related medical consultations To confirm the outbreak, the 11 GPs in Gourdon and the local hospital emergency department were requested to provide the number of total daily visits and the number of cases of gastroenteritis cases occurring between 1 August and 11 September 2000. Retrospective cohort study A retrospective cohort study was performed on a sample of Gourdon households between 5 and 15 September 2000. Domestic households were selected randomly from the telephone directory. A definite case was defined as a Gourdon resident who had at least three liquid stools ℠day or vomiting between 1 August and 3 September, and who had been present in Gourdon for ‡ 1 day between 1 and 31 August; if less than three liquid stools ℠day or abdominal pain had occurred, the case was classified as probable. Exposure was defined as the consumption of tap water before 25 August 2000. A standardised telephone questionnaire was completed to obtain demographic data (gender, date of birth), food consumption (commercially prepared dishes, sea-food, raw milk, pastry), tap water consumption (before and after 25 August, daily number of glasses of water, consumption of boiled tap water), symptoms of gastroenteritis (date of onset, number of liquid stools ℠day, vomiting, nausea, abdominal pain, fever), medical care (in- or out-patient) and consequences on daily activities (sick leave, bed confinement) for each member of the selected household. Under the hypothesis that 60% of the population consumed tap water, an estimated 550 people were required for inclusion in the study to identify a relative risk (RR) of ‡ 3, with a 5% a risk and a power of 80%. Based on an average of two individuals per household (INSERM 1999 population census), 300 households were required for inclusion in the survey. The association between AGE and the consumption of tap water was assessed by calculating the risk ratio (RR) and 95% CI, taking into account a household cluster design effect. A dose-effect relationship was evaluated by analysing the RR trend for different categories of daily number of glasses of tap water consumed (chi-square trend). Proportions were compared using the corrected chi-square test, and means were compared using the Student test or Mann and Witney test. Data were analysed with Epi-Info 6.04c (CDC, Atlanta, GA, USA). The gender and age distribution of the cohort population were compared with that of the general population of Gourdon (INSERM 1999 population census) to assess whether it was representative. Under the hypothesis of representation, the total number of cases of gastroenteritis was estimated by applying the attack rate to the total population of Gourdon, including the tourist population (data provided by the Gourdon tourism office). The most likely period for contamination of the water distribution system was based on the epidemic curve and the minimal incubation period of the pathogens involved. Microbiological analyses of stools Stool samples were examined for the presence of Salmonella spp., Shigella spp., Staphylococcus aureus, Campylobacter spp., Yersinia enterocolitica, Escherichia coli, enterococci, rotavirus groups A and C, astrovirus, calicivirus, adenovirus types 40 and 41, enterovirus, hepatitis A virus and Cryptosporidum. Viruses were detected using an enzyme immunoassay for group A rotavirus, astrovirus and adenovirus 40 ℠41, and by RT-PCR for calicivirus, group C rotavirus, enterovirus and hepatitis A virus [4–6]. Campylobacter isolates were sent to the National Reference Centre for Campylobacter for identification and fingerprinting. Species identification was by means of phenotypic tests (API Campy; bioMe´rieux, Lyon, France), with use of PCR to differentiate Campylobacter jejuni from Camplylobacter coli [7]. Random amplified polymorphic DNA fingerprinting was performed with primer 3881, 5¢-AACGCGCAAC. Environmental investigation An environmental and microbiological investigation of the water distribution system was performed. The inspection concerned the water catchments, the water distribution and sewage system, agricultural practices and private households around the sources. Bacteriological (Escherichia coli, enterococci, faecal streptococci, thermotolerant coliforms, sulphite-reducing clostridia) and chemical (chlorine) analyses were performed on tap water and on the water sampled from the two groundwater sources. After 23 August, the water chlorine concentration and the presence of coliform bacteria were monitored daily at several sites of the water system before and after the resort treatment. Examinations for viruses (group A rotavirus, astrovirus, norovirus, enterovirus, hepatitis A virus) and Campylobacter spp. were undertaken on two 10-mL tap water samples collected inhouse on 23 August, and on two 10-mL samples from groundwater sources A and B on 25 August. Viruses were detected by RT-PCR, followed by hybridisation with a specific primer to improve sensitivity, while C. jejuni and C. coli were detected by PCR with specific primers. Group A rotavirus isolates detected in stools and in water were sequenced using Beg 9 and End 9 primers [8] and compared using the Infobiogene GCG software (http://www.infobiogen.fr). Bacteriological (revived 36C, revived 22C, thermotolerant coliforms, streptococci and faecal enterococci), physical (temperature, conductivity) and chemical (oxygen consumption, ammonium, nitrates, phosphorus) parameters from source A and the river were analysed. Trends were compared for an 8-week period between 21 December 2000 and 7 February 2001. RESULTS Epidemiology Between 1 August and 11 September 2000, 7104 Gourdon residents consulted a GP or the local hospital emergency department. Between 24 and 28 August, gastroenteritis accounted for 44% (479 ℠1097) of all medical visits, compared with 6% (126 ℠2185) during the first 15 days of August (Fig. 1). Among the 498 households contacted for the household survey, 198 were excluded (103 did not respond, 49 did not fulfil the inclusion criteria, 43 refused to participate, and three for unknown reasons). The questionnaire was completed for 300 households (709 individuals). Data were missing concerning gender for one individual and age for 11 individuals. Overall, 331 (46.7%; 95% CI 44.5–49.0) of respondents were male and 377 (53.2%; 95% CI 50.9–55.5) were female. The median age was 44 years (range < 1–94 years). The age and gender distribution of the sample was similar to that of the general population. During the outbreak period, 264 of the 709 respondents (attack rate (AR) 37.2%) had been ill, with 202 definite cases (AR 28.5%; 95% CI 24.6– 32.4) and 62 probable cases (AR 8.7%; 95% CI 6.5– 11.0). The AR was higher in females (31.8%) than in males (24.8%), and was highest in children aged < 6 years (42.1%, Table 1). The epidemic curve (Fig. 2) suggests that the outbreak started between 14 and 19 August. The shape of the epidemic curve in the retrospective cohort study was very similar to that based on the number of visits to GPs for gastroenteritis. Patients complained mainly of diarrhoea, abdominal pain and nausea (Table 2). No cases of bloody diarrhoea were notified. The mean duration of symptoms was 4.5 days (median 3, range 1–30 days). Among definite cases, 52% had visited a GP, 42% had taken sick leave for ‡ 1 day, 32% had required bed rest, and 2.9% (six patients) had been hospitalised. In total, 336 (47.4%) respondents had drunk tap water before 25 August, whereas 646 (91.1%) did not drink tap water after 25 August. The risk for illness, taking into account the sample cluster design effect of 1.4, was three-fold greater (95% CI 2.2–4.1) for those who had drunk tap water before 25 August, and increased with the amount of water consumed (chi-square trend: p < 0.0001; Table 3). The risk of gastroenteritis associated with drinking tap water did not change substantially after adjustment for age (RRMH 3.3, 95% CI 2.5–4.4), and there was no modification effect according to age. Food consumption was not associated with illness. Applying the AR of 37% to the total population of Gourdon exposed to the water supply (4888 residents + 2200 tourists), the number of individuals affected by this outbreak was estimated to be 2600 (95% CI 2400–2900), assuming an exposure to tap water and a risk for gastroenteritis among the tourists similar to that among the inhabitants. Microbiological stool analyses Among the 35 stool samples analysed for bacteria, 11 (31.5%) were positive for C. coli. Twenty-four of the 35 samples were also tested for viruses: rotavirus A was detected in 17 (71.0%) and norovirus in five (21.0%) samples, while nine (37.5%) samples were co-infected (four C. coli + rotavirus; one C. coli + norovirus; three rotavirus + norovirus; and one C. coli + rotavirus + norovirus). Of the 11 C. coli isolates, six were characterised as biotype 2, and were identical following RAPD analysis, while the remaining five were of biotype 1. Rotavirus typing by RTPCR identified a single genotype P-(8), G1. Molecular characterisation of the norovirus isolates identified two different patterns, with three isolates being identical and close to the Saragota strain (genogroup I) and two close to the new GGIIb variant (genogroup II). Environmental investigation All of the water samples from source B were negative, regardless of the site of the distribution system tested. Analyses of the water originating from source A, sampled after leaving the water-works on 23 and 25 August, revealed high concentrations of E. coli (2263 / 100 mL), enterococci (415 / 100 mL), total coliforms (90 / 100 mL), faecal streptococci (136 / 100 mL) and sulphite-reducing clostridia (10 / 100 mL), and an absence of chlorination. Rotavirus A genotype G1 was detected in the crude water sampled from source A. The analyses for enterovirus, astrovirus, norovirus, hepatitis A virus and Campylobacter were negative. Source A was situated in a valley containing a river, below a hamlet of 15 dwellings, including farms breeding cattle, pork and sheep. The source was not supposed to be supplied by the river. The river resulted from the convergence of about ten different streams, with one originating close to a slaughterhouse, and was exposed to agricultural run-off and manure effluent; sheep and cattle grazed in meadows bordering the river, and a large amount of sewage from water treatment facilities was sprayed on to land close to the river. Cowsheds were close to the water catchment area. The immediate protected area around source A was too small and littered with building rubbish. Air vents without protection in the door made the source accessible to animals and insects. Bacteriological analyses of the water samples from source A and the river revealed that the contamination was of both human and animal origin, with a higher concentration in the river. The levels and trends of the chemical parameters of source A were similar to those of the river, strongly suggesting that the river was connected to source A (Fig. 3). A retrospective review of routine data collected between 1991 and 2000 showed irregularities in the maintenance of the water distribution system and the water treatment plant, particularly during the month of August. In the hamlet, the sewage system of the 15 households did not comply with sanitary rules. All but three ejected waste either into a cesspool or superficially without treatment. Three regions of gene 9 of the rotavirus detected in stools and water were compared. Sequencing showed that the strains were closely related but not identical. Assuming a likely onset of the outbreak between 14 and 19 August, and taking into account a minimum incubation period of 2 days for rotavirus A, norovirus and C. coli, the contamination of the water distribution system probably occurred between 12 and 17 August. Contamination probably persisted until 28 August, as intensive chlorination, beginning on 25 August, was effective only from 28 August onwards (Fig. 2). DISCUSSION The present descriptive and retrospective cohort studies confirmed that the outbreak of infection in the community of Gourdon was related to the consumption of tap water and affected 2600 individuals. The outbreak involved multiple pathogens, with group A rotavirus and C. coli being the most frequent, followed by norovirus. Results of environmental investigations suggested that water contamination was related to the contamination of the source, combined with a failure of the water chlorination system. The association with tap water consumption, the dose-effect relationship, and the fact that multiple pathogens were involved, were all strongly in favour of tap water as the origin of infection. Moreover, the number of cases decreased dramatically after the water restriction advice and the implementation of control measures on the water supply. No food products explored in the study were associated with illness. An AR of 14.5% and 31% was observed among adults and children aged < 6 years, respectively, who reported that they had not consumed tap water. This relatively high rate could be explained by several phenomena. Many individuals may have consumed tap water without being aware of it, e.g., in drinks or ice cubes consumed outside of the home, via raw fruits and vegetables, dishes prepared with unboiled water, or while cleaning their teeth. It can be hypothesised that people who became ill without being aware of having consumed tap water were exposed to small amounts of water and had been subjected to a small infective dose. In addition, children, more than adults, may have been contaminated through person-to-person transmission (e.g., family contact), which is the most common route of transmission for rotavirus and calicivirus. Waterborne contamination can affect large numbers of individuals [3,9–11]. In the present study, assuming a 50% increase in the population during the tourist season, the number of affected people was estimated to be between 2400 and 2900. The cohort study showed that almost half (46.5%) of the cases consulted a GP. Applying this proportion to the estimated number of cases yielded an estimated 1116–1349 ill individuals who consulted a GP for gastroenteritis, similar to the number of consultations for gastroenteritis actually reported by the GPs (n = 1037). The stool sample analyses revealed the presence of multiple pathogens (rotavirus A, norovirus genogroups I and II and C. coli) as single or co-infections. Except for C. coli, these pathogens have been implicated in numerous waterborne outbreaks, suggesting an influx of sewage into the water supply system [10,12–18]. Rotaviruses are known to be responsible for large epidemics during winter seasons, affecting particularly children aged < 3 years. These epidemics are caused by the faecal route of contamination (personto-person transmission). The high proportion of stools positive for rotavirus (71%) is uncommon in waterborne outbreaks and probably reflects massive contamination [15,16,19]. Different molecular sequences of rotavirus A genotype G1 found in stools and water indicated a human faecal source of infection. In the present study, contamination of water by sewage could be linked with the large amount of sewage from the water purification plant sprayed on to land close to the river and the water catchments. In this outbreak, C. coli, which was identified in 32% of stools, was the second most frequent pathogen. To our knowledge, human C. coli contamination has not been described previously as a causative agent in waterborne outbreaks, unlike C. jejuni [11,20–25], although C. coli has been isolated from water [26]. The main symptoms observed (diarrhoea, abdominal pain, nausea) and the mean duration of the disease (4.5 days) were compatible with a Campylobacter infection. No bloody diarrhoea was reported, although bloody diarrhoea is not constant in Campylobacter infections [27]. The high proportion of individuals who had to take sick leave, and the fact that six individuals were hospitalised, suggested a relative severity of symptoms, which is much more compatible with Campylobacter infection than with virus infections. Among the hospitalised patients, 60% were infected with C. coli. Among patients whose stools were analysed and whose age was known, cases infected only with rotavirus were younger (mean age 49 years; range, 5 months to 95 years) than cases infected only with Campylobacter (mean age 66 years; range 23–88 years) or co-infected with rotavirus plus Campylobacter (mean age 69 years; range 2–91 years). Moreover, the proportion (52%) of medical visits was unusually high compared with the proportion of visits observed during other waterborne outbreaks [13,16,28]. Campylobacter needs specific atmospheric conditions for growth and survival and its detection can be very difficult [29,30]. Because of a rapid physiological transformation to a viable, though non-cultivable, state, it has been suggested that conventional culture methods often fail [31]. Nevertheless, in the present study, the failure to detect C. coli in water, and the detection of noroviruses in stools, could be caused by the small (10 mL) water samples examined. Ha¨nninen et al. [30] proposed that the volume used for detection of suspected pathogens in drinking water should be at least 8–10 L. Groundwater is considered widely to be microbiologically clean, and is usually used for drinking without treatment. However, groundwater can be contaminated by the environment [26,32]. Stanley et al. [33] showed that Campylobacter can occur in groundwater, with hydrological evidence suggesting that the source of contamination in their study was a dairy farm situated within the hydrological catchment area of the groundwater, which was confirmed when identical strains of C. jejuni were isolated from groundwater and the dairy herd. In the present study, Campylobacter was not identified from groundwater, and dairy herds and other domestic animals in farms around the source were not investigated. However, results of bacterial, chemical and physical water analyses suggested strongly that a river contaminated by agricultural run-off from surrounding sheep and cattle grazing meadows was contaminating the source. In addition, the immediate area of the source was insufficiently protected and vulnerable to the environment. Wild birds, rodents, pigs and cows are all hosts for C. coli and could contaminate surface water easily [26,27,34,35]. Campylobacters are sensitive to many environmental factors and to chlorine [33,36,37], but a failure of the chlorination system could have resulted in the presence of a high concentration of C. coli. Waterborne outbreaks are often caused by several factors that act together [18,20,23,38]. In some outbreaks, heavy rainfall has been responsible for increasing agricultural run-off into rivers, although this did not occur before the outbreak in Gourdon. This underlies the need for increased frequency of routine water quality monitoring, chlorination of the water supply according to environmental conditions, and implementation of preventive measures. This outbreak raised the question of the delay in raising the alert. An outbreak can be recognised by: (i) the poor quality of water detected during regular controls; (ii) complaints about water quality; (iii) an incident in the supply and ℠or sewage water system during maintenance work; and (iv) an increased number of cases of gastroenteritis. In most instances, waterborne outbreaks are suspected when there are already a large number of declared cases [3,12,18,28,38,39]. In the present outbreak, the initial warning was given by a GP who notified officials 5–10 days after the assumed date of onset. This highlights the role of GPs in the detection of an unusual event, particularly during the summer, when an increase in acute gastroenteritis is less expected than during winter [40,41]. In addition, a routine water quality monitoring system able to detect contamination of drinking water before the occurrence of illness and complaints about the water quality should be used as an early warning source. The retrospective cohort study of Gourdon residents allowed the scale and severity of the outbreak to be quantified precisely and the hypothesis that tap water was the vehicle of the outbreak to be tested. When provided with a cohort sample representative of the study population, the number of people affected can be estimated. However, as the exposure of the population was high, a relatively large sample size was needed to test the association between the disease and the exposure. Because of misclassification of exposure (unrecognised consumption of tap water), the attributable risk is probably underestimated. In the absence of obvious alternative explanations, water contamination is the most likely common cause of an explosive outbreak of AGE occurring in a large population. Tillett et al. [42] proposed categorising levels of evidence for waterborne outbreaks. A descriptive study suggesting a water-related outbreak, together with the identification of the same pathogen in stools and water, should be sufficient to conclude that an outbreak is associated strongly with water consumption. Faecal indicators of water quality are currently insufficiently sensitive for some pathogens, notably viruses and parasites (e.g., Cryptosporidium). Therefore, in the absence of evidence from an analytical study to demonstrate an association, the link between illness and water can only be classified as probable or possible. However, total coliform counts are good indicators of faecal contamination with other pathogens (e.g., rotaviruses, E. coli, Campylobacter) [43]. In conclusion, the waterborne outbreak described in the present study was most probably caused by faecal contamination of source A by groundwater, resulting from deficiencies in the maintenance of the water distribution system. Corrective measures were implemented based on the results of the investigation. To our knowledge, this is the first documented waterborne outbreak involving human C. coli infection. In-depth environmental investigations to identify the origin of water contamination are important in order to better understand microbial transmission from the environment and to implement appropriate controls and preventive measures. In addition, surveys on vehicles and vectors that transmit Campylobacter between hosts are crucial in order to better understand the epidemiology of these organisms.

In which country did this happen?

288

A large multi-pathogen waterborne community outbreak linked to faecal contamination of a groundwater system, France, 2000

A large waterborne outbreak of infection that occurred during August 2000 in a local community in France was investigated initially via a rapid survey of visits to local physicians. A retrospective cohort study was then conducted on a random cluster sample of residents. Of 709 residents interviewed, 202 (28.5%) were definite cases (at least three liquid stools ℠day or vomiting) and 62 (8.7%) were probable cases (less than three liquid stools ℠day or abdominal pain). Those who had drunk tap water had a threefold increased risk for illness (95% CI 2.4–4.0). The risk increased with the amount of water consumed (chi-square trend: p < 0.0001). Bacteriological analyses of stools were performed for 35 patients and virological analyses for 24 patients. Campylobacter coli, group A rotavirus and norovirus were detected in 31.5%, 71.0% and 21% of samples, respectively. An extensive environmental investigation concluded that a groundwater source to this community had probably been contaminated by agricultural run-off, and a failure in the chlorination system was identified. This is the first documented waterborne outbreak of infection involving human C. coli infections. A better understanding of the factors influencing campylobacter transmission between hosts is required. Keywords Campylobacter coli, epidemiology, gastroenteritis, norovirus, rotavirus, waterborne outbreak INTRODUCTION Contaminated drinking water causes extensive outbreaks of illness because of the large number of people served by water supply facilities. Detection of such outbreaks requires the identification of an increase in illness (usually the rate of gastrointestinal disease) in the exposed population and confirmation that water was the route of transmission [1]. Many countries have routine reporting systems for detecting foodborne and waterborne outbreaks of infection. However, in routine practice it is not easy to detect an increase in clinical cases and to link this increase to waterborne transmission; thus waterborne outbreaks are often unrecognised and underestimated [2]. Microbiological examination of water is more complex than examination of stools, and there is often a failure to detect pathogens in water. Norovirus, Cryptosporidium parvum, Giardia intestinalis and Campylobacter jejuni are the pathogens identified most frequently [3]. The present study describes a large waterborne community outbreak of infection with multiple pathogens, including Campylobacter coli. METHODS AND MATERIALS Background On 23 August 2000, a general practitioner (GP) informed the local district health department that 16 cases of acute gastroenteritis (AGE) had occurred among residents of a holiday camp in the Gourdon community since 21 August. Gourdon, which is located in the south-west of France, has a population of 4888 inhabitants, which increases by 50% in the July– August tourist season. Investigations at the holiday camp concluded that foodborne transmission was very unlikely, but found that tap water was highly contaminated with faecal coliforms. Preliminary information collected from physicians in Gourdon indicated an increase of the number of consultations for AGE. Gourdon tap water is supplied by two underground water facilities: facility A serving 3800 water consumers in Gourdon; and facility B serving 1200 water consumers in Gourdon and the surrounding communities. Water treatment in facility A consists of chlorination, which involves two automatic pumps working in rotation. The water then passes through two semi in-ground reservoir tanks (1100 m3 and 200 m3, respectively) before distribution. Facility B water is mixed with a fraction of facility A water in another 500 m3 semi in-ground reservoir tank before being distributed. This mixing with the treated water from facility A is intended to disinfect the water from facility B. Since the increase in AGE was noted only in Gourdon, contaminated tap water from facility A was suspected to be the origin of the outbreak. Therefore, the population of Gourdon was informed on 25 August to avoid the consumption of tap water or to boil it for 5 min. Survey of gastroenteritis-related medical consultations To confirm the outbreak, the 11 GPs in Gourdon and the local hospital emergency department were requested to provide the number of total daily visits and the number of cases of gastroenteritis cases occurring between 1 August and 11 September 2000. Retrospective cohort study A retrospective cohort study was performed on a sample of Gourdon households between 5 and 15 September 2000. Domestic households were selected randomly from the telephone directory. A definite case was defined as a Gourdon resident who had at least three liquid stools ℠day or vomiting between 1 August and 3 September, and who had been present in Gourdon for ‡ 1 day between 1 and 31 August; if less than three liquid stools ℠day or abdominal pain had occurred, the case was classified as probable. Exposure was defined as the consumption of tap water before 25 August 2000. A standardised telephone questionnaire was completed to obtain demographic data (gender, date of birth), food consumption (commercially prepared dishes, sea-food, raw milk, pastry), tap water consumption (before and after 25 August, daily number of glasses of water, consumption of boiled tap water), symptoms of gastroenteritis (date of onset, number of liquid stools ℠day, vomiting, nausea, abdominal pain, fever), medical care (in- or out-patient) and consequences on daily activities (sick leave, bed confinement) for each member of the selected household. Under the hypothesis that 60% of the population consumed tap water, an estimated 550 people were required for inclusion in the study to identify a relative risk (RR) of ‡ 3, with a 5% a risk and a power of 80%. Based on an average of two individuals per household (INSERM 1999 population census), 300 households were required for inclusion in the survey. The association between AGE and the consumption of tap water was assessed by calculating the risk ratio (RR) and 95% CI, taking into account a household cluster design effect. A dose-effect relationship was evaluated by analysing the RR trend for different categories of daily number of glasses of tap water consumed (chi-square trend). Proportions were compared using the corrected chi-square test, and means were compared using the Student test or Mann and Witney test. Data were analysed with Epi-Info 6.04c (CDC, Atlanta, GA, USA). The gender and age distribution of the cohort population were compared with that of the general population of Gourdon (INSERM 1999 population census) to assess whether it was representative. Under the hypothesis of representation, the total number of cases of gastroenteritis was estimated by applying the attack rate to the total population of Gourdon, including the tourist population (data provided by the Gourdon tourism office). The most likely period for contamination of the water distribution system was based on the epidemic curve and the minimal incubation period of the pathogens involved. Microbiological analyses of stools Stool samples were examined for the presence of Salmonella spp., Shigella spp., Staphylococcus aureus, Campylobacter spp., Yersinia enterocolitica, Escherichia coli, enterococci, rotavirus groups A and C, astrovirus, calicivirus, adenovirus types 40 and 41, enterovirus, hepatitis A virus and Cryptosporidum. Viruses were detected using an enzyme immunoassay for group A rotavirus, astrovirus and adenovirus 40 ℠41, and by RT-PCR for calicivirus, group C rotavirus, enterovirus and hepatitis A virus [4–6]. Campylobacter isolates were sent to the National Reference Centre for Campylobacter for identification and fingerprinting. Species identification was by means of phenotypic tests (API Campy; bioMe´rieux, Lyon, France), with use of PCR to differentiate Campylobacter jejuni from Camplylobacter coli [7]. Random amplified polymorphic DNA fingerprinting was performed with primer 3881, 5¢-AACGCGCAAC. Environmental investigation An environmental and microbiological investigation of the water distribution system was performed. The inspection concerned the water catchments, the water distribution and sewage system, agricultural practices and private households around the sources. Bacteriological (Escherichia coli, enterococci, faecal streptococci, thermotolerant coliforms, sulphite-reducing clostridia) and chemical (chlorine) analyses were performed on tap water and on the water sampled from the two groundwater sources. After 23 August, the water chlorine concentration and the presence of coliform bacteria were monitored daily at several sites of the water system before and after the resort treatment. Examinations for viruses (group A rotavirus, astrovirus, norovirus, enterovirus, hepatitis A virus) and Campylobacter spp. were undertaken on two 10-mL tap water samples collected inhouse on 23 August, and on two 10-mL samples from groundwater sources A and B on 25 August. Viruses were detected by RT-PCR, followed by hybridisation with a specific primer to improve sensitivity, while C. jejuni and C. coli were detected by PCR with specific primers. Group A rotavirus isolates detected in stools and in water were sequenced using Beg 9 and End 9 primers [8] and compared using the Infobiogene GCG software (http://www.infobiogen.fr). Bacteriological (revived 36C, revived 22C, thermotolerant coliforms, streptococci and faecal enterococci), physical (temperature, conductivity) and chemical (oxygen consumption, ammonium, nitrates, phosphorus) parameters from source A and the river were analysed. Trends were compared for an 8-week period between 21 December 2000 and 7 February 2001. RESULTS Epidemiology Between 1 August and 11 September 2000, 7104 Gourdon residents consulted a GP or the local hospital emergency department. Between 24 and 28 August, gastroenteritis accounted for 44% (479 ℠1097) of all medical visits, compared with 6% (126 ℠2185) during the first 15 days of August (Fig. 1). Among the 498 households contacted for the household survey, 198 were excluded (103 did not respond, 49 did not fulfil the inclusion criteria, 43 refused to participate, and three for unknown reasons). The questionnaire was completed for 300 households (709 individuals). Data were missing concerning gender for one individual and age for 11 individuals. Overall, 331 (46.7%; 95% CI 44.5–49.0) of respondents were male and 377 (53.2%; 95% CI 50.9–55.5) were female. The median age was 44 years (range < 1–94 years). The age and gender distribution of the sample was similar to that of the general population. During the outbreak period, 264 of the 709 respondents (attack rate (AR) 37.2%) had been ill, with 202 definite cases (AR 28.5%; 95% CI 24.6– 32.4) and 62 probable cases (AR 8.7%; 95% CI 6.5– 11.0). The AR was higher in females (31.8%) than in males (24.8%), and was highest in children aged < 6 years (42.1%, Table 1). The epidemic curve (Fig. 2) suggests that the outbreak started between 14 and 19 August. The shape of the epidemic curve in the retrospective cohort study was very similar to that based on the number of visits to GPs for gastroenteritis. Patients complained mainly of diarrhoea, abdominal pain and nausea (Table 2). No cases of bloody diarrhoea were notified. The mean duration of symptoms was 4.5 days (median 3, range 1–30 days). Among definite cases, 52% had visited a GP, 42% had taken sick leave for ‡ 1 day, 32% had required bed rest, and 2.9% (six patients) had been hospitalised. In total, 336 (47.4%) respondents had drunk tap water before 25 August, whereas 646 (91.1%) did not drink tap water after 25 August. The risk for illness, taking into account the sample cluster design effect of 1.4, was three-fold greater (95% CI 2.2–4.1) for those who had drunk tap water before 25 August, and increased with the amount of water consumed (chi-square trend: p < 0.0001; Table 3). The risk of gastroenteritis associated with drinking tap water did not change substantially after adjustment for age (RRMH 3.3, 95% CI 2.5–4.4), and there was no modification effect according to age. Food consumption was not associated with illness. Applying the AR of 37% to the total population of Gourdon exposed to the water supply (4888 residents + 2200 tourists), the number of individuals affected by this outbreak was estimated to be 2600 (95% CI 2400–2900), assuming an exposure to tap water and a risk for gastroenteritis among the tourists similar to that among the inhabitants. Microbiological stool analyses Among the 35 stool samples analysed for bacteria, 11 (31.5%) were positive for C. coli. Twenty-four of the 35 samples were also tested for viruses: rotavirus A was detected in 17 (71.0%) and norovirus in five (21.0%) samples, while nine (37.5%) samples were co-infected (four C. coli + rotavirus; one C. coli + norovirus; three rotavirus + norovirus; and one C. coli + rotavirus + norovirus). Of the 11 C. coli isolates, six were characterised as biotype 2, and were identical following RAPD analysis, while the remaining five were of biotype 1. Rotavirus typing by RTPCR identified a single genotype P-(8), G1. Molecular characterisation of the norovirus isolates identified two different patterns, with three isolates being identical and close to the Saragota strain (genogroup I) and two close to the new GGIIb variant (genogroup II). Environmental investigation All of the water samples from source B were negative, regardless of the site of the distribution system tested. Analyses of the water originating from source A, sampled after leaving the water-works on 23 and 25 August, revealed high concentrations of E. coli (2263 / 100 mL), enterococci (415 / 100 mL), total coliforms (90 / 100 mL), faecal streptococci (136 / 100 mL) and sulphite-reducing clostridia (10 / 100 mL), and an absence of chlorination. Rotavirus A genotype G1 was detected in the crude water sampled from source A. The analyses for enterovirus, astrovirus, norovirus, hepatitis A virus and Campylobacter were negative. Source A was situated in a valley containing a river, below a hamlet of 15 dwellings, including farms breeding cattle, pork and sheep. The source was not supposed to be supplied by the river. The river resulted from the convergence of about ten different streams, with one originating close to a slaughterhouse, and was exposed to agricultural run-off and manure effluent; sheep and cattle grazed in meadows bordering the river, and a large amount of sewage from water treatment facilities was sprayed on to land close to the river. Cowsheds were close to the water catchment area. The immediate protected area around source A was too small and littered with building rubbish. Air vents without protection in the door made the source accessible to animals and insects. Bacteriological analyses of the water samples from source A and the river revealed that the contamination was of both human and animal origin, with a higher concentration in the river. The levels and trends of the chemical parameters of source A were similar to those of the river, strongly suggesting that the river was connected to source A (Fig. 3). A retrospective review of routine data collected between 1991 and 2000 showed irregularities in the maintenance of the water distribution system and the water treatment plant, particularly during the month of August. In the hamlet, the sewage system of the 15 households did not comply with sanitary rules. All but three ejected waste either into a cesspool or superficially without treatment. Three regions of gene 9 of the rotavirus detected in stools and water were compared. Sequencing showed that the strains were closely related but not identical. Assuming a likely onset of the outbreak between 14 and 19 August, and taking into account a minimum incubation period of 2 days for rotavirus A, norovirus and C. coli, the contamination of the water distribution system probably occurred between 12 and 17 August. Contamination probably persisted until 28 August, as intensive chlorination, beginning on 25 August, was effective only from 28 August onwards (Fig. 2). DISCUSSION The present descriptive and retrospective cohort studies confirmed that the outbreak of infection in the community of Gourdon was related to the consumption of tap water and affected 2600 individuals. The outbreak involved multiple pathogens, with group A rotavirus and C. coli being the most frequent, followed by norovirus. Results of environmental investigations suggested that water contamination was related to the contamination of the source, combined with a failure of the water chlorination system. The association with tap water consumption, the dose-effect relationship, and the fact that multiple pathogens were involved, were all strongly in favour of tap water as the origin of infection. Moreover, the number of cases decreased dramatically after the water restriction advice and the implementation of control measures on the water supply. No food products explored in the study were associated with illness. An AR of 14.5% and 31% was observed among adults and children aged < 6 years, respectively, who reported that they had not consumed tap water. This relatively high rate could be explained by several phenomena. Many individuals may have consumed tap water without being aware of it, e.g., in drinks or ice cubes consumed outside of the home, via raw fruits and vegetables, dishes prepared with unboiled water, or while cleaning their teeth. It can be hypothesised that people who became ill without being aware of having consumed tap water were exposed to small amounts of water and had been subjected to a small infective dose. In addition, children, more than adults, may have been contaminated through person-to-person transmission (e.g., family contact), which is the most common route of transmission for rotavirus and calicivirus. Waterborne contamination can affect large numbers of individuals [3,9–11]. In the present study, assuming a 50% increase in the population during the tourist season, the number of affected people was estimated to be between 2400 and 2900. The cohort study showed that almost half (46.5%) of the cases consulted a GP. Applying this proportion to the estimated number of cases yielded an estimated 1116–1349 ill individuals who consulted a GP for gastroenteritis, similar to the number of consultations for gastroenteritis actually reported by the GPs (n = 1037). The stool sample analyses revealed the presence of multiple pathogens (rotavirus A, norovirus genogroups I and II and C. coli) as single or co-infections. Except for C. coli, these pathogens have been implicated in numerous waterborne outbreaks, suggesting an influx of sewage into the water supply system [10,12–18]. Rotaviruses are known to be responsible for large epidemics during winter seasons, affecting particularly children aged < 3 years. These epidemics are caused by the faecal route of contamination (personto-person transmission). The high proportion of stools positive for rotavirus (71%) is uncommon in waterborne outbreaks and probably reflects massive contamination [15,16,19]. Different molecular sequences of rotavirus A genotype G1 found in stools and water indicated a human faecal source of infection. In the present study, contamination of water by sewage could be linked with the large amount of sewage from the water purification plant sprayed on to land close to the river and the water catchments. In this outbreak, C. coli, which was identified in 32% of stools, was the second most frequent pathogen. To our knowledge, human C. coli contamination has not been described previously as a causative agent in waterborne outbreaks, unlike C. jejuni [11,20–25], although C. coli has been isolated from water [26]. The main symptoms observed (diarrhoea, abdominal pain, nausea) and the mean duration of the disease (4.5 days) were compatible with a Campylobacter infection. No bloody diarrhoea was reported, although bloody diarrhoea is not constant in Campylobacter infections [27]. The high proportion of individuals who had to take sick leave, and the fact that six individuals were hospitalised, suggested a relative severity of symptoms, which is much more compatible with Campylobacter infection than with virus infections. Among the hospitalised patients, 60% were infected with C. coli. Among patients whose stools were analysed and whose age was known, cases infected only with rotavirus were younger (mean age 49 years; range, 5 months to 95 years) than cases infected only with Campylobacter (mean age 66 years; range 23–88 years) or co-infected with rotavirus plus Campylobacter (mean age 69 years; range 2–91 years). Moreover, the proportion (52%) of medical visits was unusually high compared with the proportion of visits observed during other waterborne outbreaks [13,16,28]. Campylobacter needs specific atmospheric conditions for growth and survival and its detection can be very difficult [29,30]. Because of a rapid physiological transformation to a viable, though non-cultivable, state, it has been suggested that conventional culture methods often fail [31]. Nevertheless, in the present study, the failure to detect C. coli in water, and the detection of noroviruses in stools, could be caused by the small (10 mL) water samples examined. Ha¨nninen et al. [30] proposed that the volume used for detection of suspected pathogens in drinking water should be at least 8–10 L. Groundwater is considered widely to be microbiologically clean, and is usually used for drinking without treatment. However, groundwater can be contaminated by the environment [26,32]. Stanley et al. [33] showed that Campylobacter can occur in groundwater, with hydrological evidence suggesting that the source of contamination in their study was a dairy farm situated within the hydrological catchment area of the groundwater, which was confirmed when identical strains of C. jejuni were isolated from groundwater and the dairy herd. In the present study, Campylobacter was not identified from groundwater, and dairy herds and other domestic animals in farms around the source were not investigated. However, results of bacterial, chemical and physical water analyses suggested strongly that a river contaminated by agricultural run-off from surrounding sheep and cattle grazing meadows was contaminating the source. In addition, the immediate area of the source was insufficiently protected and vulnerable to the environment. Wild birds, rodents, pigs and cows are all hosts for C. coli and could contaminate surface water easily [26,27,34,35]. Campylobacters are sensitive to many environmental factors and to chlorine [33,36,37], but a failure of the chlorination system could have resulted in the presence of a high concentration of C. coli. Waterborne outbreaks are often caused by several factors that act together [18,20,23,38]. In some outbreaks, heavy rainfall has been responsible for increasing agricultural run-off into rivers, although this did not occur before the outbreak in Gourdon. This underlies the need for increased frequency of routine water quality monitoring, chlorination of the water supply according to environmental conditions, and implementation of preventive measures. This outbreak raised the question of the delay in raising the alert. An outbreak can be recognised by: (i) the poor quality of water detected during regular controls; (ii) complaints about water quality; (iii) an incident in the supply and ℠or sewage water system during maintenance work; and (iv) an increased number of cases of gastroenteritis. In most instances, waterborne outbreaks are suspected when there are already a large number of declared cases [3,12,18,28,38,39]. In the present outbreak, the initial warning was given by a GP who notified officials 5–10 days after the assumed date of onset. This highlights the role of GPs in the detection of an unusual event, particularly during the summer, when an increase in acute gastroenteritis is less expected than during winter [40,41]. In addition, a routine water quality monitoring system able to detect contamination of drinking water before the occurrence of illness and complaints about the water quality should be used as an early warning source. The retrospective cohort study of Gourdon residents allowed the scale and severity of the outbreak to be quantified precisely and the hypothesis that tap water was the vehicle of the outbreak to be tested. When provided with a cohort sample representative of the study population, the number of people affected can be estimated. However, as the exposure of the population was high, a relatively large sample size was needed to test the association between the disease and the exposure. Because of misclassification of exposure (unrecognised consumption of tap water), the attributable risk is probably underestimated. In the absence of obvious alternative explanations, water contamination is the most likely common cause of an explosive outbreak of AGE occurring in a large population. Tillett et al. [42] proposed categorising levels of evidence for waterborne outbreaks. A descriptive study suggesting a water-related outbreak, together with the identification of the same pathogen in stools and water, should be sufficient to conclude that an outbreak is associated strongly with water consumption. Faecal indicators of water quality are currently insufficiently sensitive for some pathogens, notably viruses and parasites (e.g., Cryptosporidium). Therefore, in the absence of evidence from an analytical study to demonstrate an association, the link between illness and water can only be classified as probable or possible. However, total coliform counts are good indicators of faecal contamination with other pathogens (e.g., rotaviruses, E. coli, Campylobacter) [43]. In conclusion, the waterborne outbreak described in the present study was most probably caused by faecal contamination of source A by groundwater, resulting from deficiencies in the maintenance of the water distribution system. Corrective measures were implemented based on the results of the investigation. To our knowledge, this is the first documented waterborne outbreak involving human C. coli infection. In-depth environmental investigations to identify the origin of water contamination are important in order to better understand microbial transmission from the environment and to implement appropriate controls and preventive measures. In addition, surveys on vehicles and vectors that transmit Campylobacter between hosts are crucial in order to better understand the epidemiology of these organisms.

Where did this happen?

289

A large multi-pathogen waterborne community outbreak linked to faecal contamination of a groundwater system, France, 2000

A large waterborne outbreak of infection that occurred during August 2000 in a local community in France was investigated initially via a rapid survey of visits to local physicians. A retrospective cohort study was then conducted on a random cluster sample of residents. Of 709 residents interviewed, 202 (28.5%) were definite cases (at least three liquid stools ℠day or vomiting) and 62 (8.7%) were probable cases (less than three liquid stools ℠day or abdominal pain). Those who had drunk tap water had a threefold increased risk for illness (95% CI 2.4–4.0). The risk increased with the amount of water consumed (chi-square trend: p < 0.0001). Bacteriological analyses of stools were performed for 35 patients and virological analyses for 24 patients. Campylobacter coli, group A rotavirus and norovirus were detected in 31.5%, 71.0% and 21% of samples, respectively. An extensive environmental investigation concluded that a groundwater source to this community had probably been contaminated by agricultural run-off, and a failure in the chlorination system was identified. This is the first documented waterborne outbreak of infection involving human C. coli infections. A better understanding of the factors influencing campylobacter transmission between hosts is required. Keywords Campylobacter coli, epidemiology, gastroenteritis, norovirus, rotavirus, waterborne outbreak INTRODUCTION Contaminated drinking water causes extensive outbreaks of illness because of the large number of people served by water supply facilities. Detection of such outbreaks requires the identification of an increase in illness (usually the rate of gastrointestinal disease) in the exposed population and confirmation that water was the route of transmission [1]. Many countries have routine reporting systems for detecting foodborne and waterborne outbreaks of infection. However, in routine practice it is not easy to detect an increase in clinical cases and to link this increase to waterborne transmission; thus waterborne outbreaks are often unrecognised and underestimated [2]. Microbiological examination of water is more complex than examination of stools, and there is often a failure to detect pathogens in water. Norovirus, Cryptosporidium parvum, Giardia intestinalis and Campylobacter jejuni are the pathogens identified most frequently [3]. The present study describes a large waterborne community outbreak of infection with multiple pathogens, including Campylobacter coli. METHODS AND MATERIALS Background On 23 August 2000, a general practitioner (GP) informed the local district health department that 16 cases of acute gastroenteritis (AGE) had occurred among residents of a holiday camp in the Gourdon community since 21 August. Gourdon, which is located in the south-west of France, has a population of 4888 inhabitants, which increases by 50% in the July– August tourist season. Investigations at the holiday camp concluded that foodborne transmission was very unlikely, but found that tap water was highly contaminated with faecal coliforms. Preliminary information collected from physicians in Gourdon indicated an increase of the number of consultations for AGE. Gourdon tap water is supplied by two underground water facilities: facility A serving 3800 water consumers in Gourdon; and facility B serving 1200 water consumers in Gourdon and the surrounding communities. Water treatment in facility A consists of chlorination, which involves two automatic pumps working in rotation. The water then passes through two semi in-ground reservoir tanks (1100 m3 and 200 m3, respectively) before distribution. Facility B water is mixed with a fraction of facility A water in another 500 m3 semi in-ground reservoir tank before being distributed. This mixing with the treated water from facility A is intended to disinfect the water from facility B. Since the increase in AGE was noted only in Gourdon, contaminated tap water from facility A was suspected to be the origin of the outbreak. Therefore, the population of Gourdon was informed on 25 August to avoid the consumption of tap water or to boil it for 5 min. Survey of gastroenteritis-related medical consultations To confirm the outbreak, the 11 GPs in Gourdon and the local hospital emergency department were requested to provide the number of total daily visits and the number of cases of gastroenteritis cases occurring between 1 August and 11 September 2000. Retrospective cohort study A retrospective cohort study was performed on a sample of Gourdon households between 5 and 15 September 2000. Domestic households were selected randomly from the telephone directory. A definite case was defined as a Gourdon resident who had at least three liquid stools ℠day or vomiting between 1 August and 3 September, and who had been present in Gourdon for ‡ 1 day between 1 and 31 August; if less than three liquid stools ℠day or abdominal pain had occurred, the case was classified as probable. Exposure was defined as the consumption of tap water before 25 August 2000. A standardised telephone questionnaire was completed to obtain demographic data (gender, date of birth), food consumption (commercially prepared dishes, sea-food, raw milk, pastry), tap water consumption (before and after 25 August, daily number of glasses of water, consumption of boiled tap water), symptoms of gastroenteritis (date of onset, number of liquid stools ℠day, vomiting, nausea, abdominal pain, fever), medical care (in- or out-patient) and consequences on daily activities (sick leave, bed confinement) for each member of the selected household. Under the hypothesis that 60% of the population consumed tap water, an estimated 550 people were required for inclusion in the study to identify a relative risk (RR) of ‡ 3, with a 5% a risk and a power of 80%. Based on an average of two individuals per household (INSERM 1999 population census), 300 households were required for inclusion in the survey. The association between AGE and the consumption of tap water was assessed by calculating the risk ratio (RR) and 95% CI, taking into account a household cluster design effect. A dose-effect relationship was evaluated by analysing the RR trend for different categories of daily number of glasses of tap water consumed (chi-square trend). Proportions were compared using the corrected chi-square test, and means were compared using the Student test or Mann and Witney test. Data were analysed with Epi-Info 6.04c (CDC, Atlanta, GA, USA). The gender and age distribution of the cohort population were compared with that of the general population of Gourdon (INSERM 1999 population census) to assess whether it was representative. Under the hypothesis of representation, the total number of cases of gastroenteritis was estimated by applying the attack rate to the total population of Gourdon, including the tourist population (data provided by the Gourdon tourism office). The most likely period for contamination of the water distribution system was based on the epidemic curve and the minimal incubation period of the pathogens involved. Microbiological analyses of stools Stool samples were examined for the presence of Salmonella spp., Shigella spp., Staphylococcus aureus, Campylobacter spp., Yersinia enterocolitica, Escherichia coli, enterococci, rotavirus groups A and C, astrovirus, calicivirus, adenovirus types 40 and 41, enterovirus, hepatitis A virus and Cryptosporidum. Viruses were detected using an enzyme immunoassay for group A rotavirus, astrovirus and adenovirus 40 ℠41, and by RT-PCR for calicivirus, group C rotavirus, enterovirus and hepatitis A virus [4–6]. Campylobacter isolates were sent to the National Reference Centre for Campylobacter for identification and fingerprinting. Species identification was by means of phenotypic tests (API Campy; bioMe´rieux, Lyon, France), with use of PCR to differentiate Campylobacter jejuni from Camplylobacter coli [7]. Random amplified polymorphic DNA fingerprinting was performed with primer 3881, 5¢-AACGCGCAAC. Environmental investigation An environmental and microbiological investigation of the water distribution system was performed. The inspection concerned the water catchments, the water distribution and sewage system, agricultural practices and private households around the sources. Bacteriological (Escherichia coli, enterococci, faecal streptococci, thermotolerant coliforms, sulphite-reducing clostridia) and chemical (chlorine) analyses were performed on tap water and on the water sampled from the two groundwater sources. After 23 August, the water chlorine concentration and the presence of coliform bacteria were monitored daily at several sites of the water system before and after the resort treatment. Examinations for viruses (group A rotavirus, astrovirus, norovirus, enterovirus, hepatitis A virus) and Campylobacter spp. were undertaken on two 10-mL tap water samples collected inhouse on 23 August, and on two 10-mL samples from groundwater sources A and B on 25 August. Viruses were detected by RT-PCR, followed by hybridisation with a specific primer to improve sensitivity, while C. jejuni and C. coli were detected by PCR with specific primers. Group A rotavirus isolates detected in stools and in water were sequenced using Beg 9 and End 9 primers [8] and compared using the Infobiogene GCG software (http://www.infobiogen.fr). Bacteriological (revived 36C, revived 22C, thermotolerant coliforms, streptococci and faecal enterococci), physical (temperature, conductivity) and chemical (oxygen consumption, ammonium, nitrates, phosphorus) parameters from source A and the river were analysed. Trends were compared for an 8-week period between 21 December 2000 and 7 February 2001. RESULTS Epidemiology Between 1 August and 11 September 2000, 7104 Gourdon residents consulted a GP or the local hospital emergency department. Between 24 and 28 August, gastroenteritis accounted for 44% (479 ℠1097) of all medical visits, compared with 6% (126 ℠2185) during the first 15 days of August (Fig. 1). Among the 498 households contacted for the household survey, 198 were excluded (103 did not respond, 49 did not fulfil the inclusion criteria, 43 refused to participate, and three for unknown reasons). The questionnaire was completed for 300 households (709 individuals). Data were missing concerning gender for one individual and age for 11 individuals. Overall, 331 (46.7%; 95% CI 44.5–49.0) of respondents were male and 377 (53.2%; 95% CI 50.9–55.5) were female. The median age was 44 years (range < 1–94 years). The age and gender distribution of the sample was similar to that of the general population. During the outbreak period, 264 of the 709 respondents (attack rate (AR) 37.2%) had been ill, with 202 definite cases (AR 28.5%; 95% CI 24.6– 32.4) and 62 probable cases (AR 8.7%; 95% CI 6.5– 11.0). The AR was higher in females (31.8%) than in males (24.8%), and was highest in children aged < 6 years (42.1%, Table 1). The epidemic curve (Fig. 2) suggests that the outbreak started between 14 and 19 August. The shape of the epidemic curve in the retrospective cohort study was very similar to that based on the number of visits to GPs for gastroenteritis. Patients complained mainly of diarrhoea, abdominal pain and nausea (Table 2). No cases of bloody diarrhoea were notified. The mean duration of symptoms was 4.5 days (median 3, range 1–30 days). Among definite cases, 52% had visited a GP, 42% had taken sick leave for ‡ 1 day, 32% had required bed rest, and 2.9% (six patients) had been hospitalised. In total, 336 (47.4%) respondents had drunk tap water before 25 August, whereas 646 (91.1%) did not drink tap water after 25 August. The risk for illness, taking into account the sample cluster design effect of 1.4, was three-fold greater (95% CI 2.2–4.1) for those who had drunk tap water before 25 August, and increased with the amount of water consumed (chi-square trend: p < 0.0001; Table 3). The risk of gastroenteritis associated with drinking tap water did not change substantially after adjustment for age (RRMH 3.3, 95% CI 2.5–4.4), and there was no modification effect according to age. Food consumption was not associated with illness. Applying the AR of 37% to the total population of Gourdon exposed to the water supply (4888 residents + 2200 tourists), the number of individuals affected by this outbreak was estimated to be 2600 (95% CI 2400–2900), assuming an exposure to tap water and a risk for gastroenteritis among the tourists similar to that among the inhabitants. Microbiological stool analyses Among the 35 stool samples analysed for bacteria, 11 (31.5%) were positive for C. coli. Twenty-four of the 35 samples were also tested for viruses: rotavirus A was detected in 17 (71.0%) and norovirus in five (21.0%) samples, while nine (37.5%) samples were co-infected (four C. coli + rotavirus; one C. coli + norovirus; three rotavirus + norovirus; and one C. coli + rotavirus + norovirus). Of the 11 C. coli isolates, six were characterised as biotype 2, and were identical following RAPD analysis, while the remaining five were of biotype 1. Rotavirus typing by RTPCR identified a single genotype P-(8), G1. Molecular characterisation of the norovirus isolates identified two different patterns, with three isolates being identical and close to the Saragota strain (genogroup I) and two close to the new GGIIb variant (genogroup II). Environmental investigation All of the water samples from source B were negative, regardless of the site of the distribution system tested. Analyses of the water originating from source A, sampled after leaving the water-works on 23 and 25 August, revealed high concentrations of E. coli (2263 / 100 mL), enterococci (415 / 100 mL), total coliforms (90 / 100 mL), faecal streptococci (136 / 100 mL) and sulphite-reducing clostridia (10 / 100 mL), and an absence of chlorination. Rotavirus A genotype G1 was detected in the crude water sampled from source A. The analyses for enterovirus, astrovirus, norovirus, hepatitis A virus and Campylobacter were negative. Source A was situated in a valley containing a river, below a hamlet of 15 dwellings, including farms breeding cattle, pork and sheep. The source was not supposed to be supplied by the river. The river resulted from the convergence of about ten different streams, with one originating close to a slaughterhouse, and was exposed to agricultural run-off and manure effluent; sheep and cattle grazed in meadows bordering the river, and a large amount of sewage from water treatment facilities was sprayed on to land close to the river. Cowsheds were close to the water catchment area. The immediate protected area around source A was too small and littered with building rubbish. Air vents without protection in the door made the source accessible to animals and insects. Bacteriological analyses of the water samples from source A and the river revealed that the contamination was of both human and animal origin, with a higher concentration in the river. The levels and trends of the chemical parameters of source A were similar to those of the river, strongly suggesting that the river was connected to source A (Fig. 3). A retrospective review of routine data collected between 1991 and 2000 showed irregularities in the maintenance of the water distribution system and the water treatment plant, particularly during the month of August. In the hamlet, the sewage system of the 15 households did not comply with sanitary rules. All but three ejected waste either into a cesspool or superficially without treatment. Three regions of gene 9 of the rotavirus detected in stools and water were compared. Sequencing showed that the strains were closely related but not identical. Assuming a likely onset of the outbreak between 14 and 19 August, and taking into account a minimum incubation period of 2 days for rotavirus A, norovirus and C. coli, the contamination of the water distribution system probably occurred between 12 and 17 August. Contamination probably persisted until 28 August, as intensive chlorination, beginning on 25 August, was effective only from 28 August onwards (Fig. 2). DISCUSSION The present descriptive and retrospective cohort studies confirmed that the outbreak of infection in the community of Gourdon was related to the consumption of tap water and affected 2600 individuals. The outbreak involved multiple pathogens, with group A rotavirus and C. coli being the most frequent, followed by norovirus. Results of environmental investigations suggested that water contamination was related to the contamination of the source, combined with a failure of the water chlorination system. The association with tap water consumption, the dose-effect relationship, and the fact that multiple pathogens were involved, were all strongly in favour of tap water as the origin of infection. Moreover, the number of cases decreased dramatically after the water restriction advice and the implementation of control measures on the water supply. No food products explored in the study were associated with illness. An AR of 14.5% and 31% was observed among adults and children aged < 6 years, respectively, who reported that they had not consumed tap water. This relatively high rate could be explained by several phenomena. Many individuals may have consumed tap water without being aware of it, e.g., in drinks or ice cubes consumed outside of the home, via raw fruits and vegetables, dishes prepared with unboiled water, or while cleaning their teeth. It can be hypothesised that people who became ill without being aware of having consumed tap water were exposed to small amounts of water and had been subjected to a small infective dose. In addition, children, more than adults, may have been contaminated through person-to-person transmission (e.g., family contact), which is the most common route of transmission for rotavirus and calicivirus. Waterborne contamination can affect large numbers of individuals [3,9–11]. In the present study, assuming a 50% increase in the population during the tourist season, the number of affected people was estimated to be between 2400 and 2900. The cohort study showed that almost half (46.5%) of the cases consulted a GP. Applying this proportion to the estimated number of cases yielded an estimated 1116–1349 ill individuals who consulted a GP for gastroenteritis, similar to the number of consultations for gastroenteritis actually reported by the GPs (n = 1037). The stool sample analyses revealed the presence of multiple pathogens (rotavirus A, norovirus genogroups I and II and C. coli) as single or co-infections. Except for C. coli, these pathogens have been implicated in numerous waterborne outbreaks, suggesting an influx of sewage into the water supply system [10,12–18]. Rotaviruses are known to be responsible for large epidemics during winter seasons, affecting particularly children aged < 3 years. These epidemics are caused by the faecal route of contamination (personto-person transmission). The high proportion of stools positive for rotavirus (71%) is uncommon in waterborne outbreaks and probably reflects massive contamination [15,16,19]. Different molecular sequences of rotavirus A genotype G1 found in stools and water indicated a human faecal source of infection. In the present study, contamination of water by sewage could be linked with the large amount of sewage from the water purification plant sprayed on to land close to the river and the water catchments. In this outbreak, C. coli, which was identified in 32% of stools, was the second most frequent pathogen. To our knowledge, human C. coli contamination has not been described previously as a causative agent in waterborne outbreaks, unlike C. jejuni [11,20–25], although C. coli has been isolated from water [26]. The main symptoms observed (diarrhoea, abdominal pain, nausea) and the mean duration of the disease (4.5 days) were compatible with a Campylobacter infection. No bloody diarrhoea was reported, although bloody diarrhoea is not constant in Campylobacter infections [27]. The high proportion of individuals who had to take sick leave, and the fact that six individuals were hospitalised, suggested a relative severity of symptoms, which is much more compatible with Campylobacter infection than with virus infections. Among the hospitalised patients, 60% were infected with C. coli. Among patients whose stools were analysed and whose age was known, cases infected only with rotavirus were younger (mean age 49 years; range, 5 months to 95 years) than cases infected only with Campylobacter (mean age 66 years; range 23–88 years) or co-infected with rotavirus plus Campylobacter (mean age 69 years; range 2–91 years). Moreover, the proportion (52%) of medical visits was unusually high compared with the proportion of visits observed during other waterborne outbreaks [13,16,28]. Campylobacter needs specific atmospheric conditions for growth and survival and its detection can be very difficult [29,30]. Because of a rapid physiological transformation to a viable, though non-cultivable, state, it has been suggested that conventional culture methods often fail [31]. Nevertheless, in the present study, the failure to detect C. coli in water, and the detection of noroviruses in stools, could be caused by the small (10 mL) water samples examined. Ha¨nninen et al. [30] proposed that the volume used for detection of suspected pathogens in drinking water should be at least 8–10 L. Groundwater is considered widely to be microbiologically clean, and is usually used for drinking without treatment. However, groundwater can be contaminated by the environment [26,32]. Stanley et al. [33] showed that Campylobacter can occur in groundwater, with hydrological evidence suggesting that the source of contamination in their study was a dairy farm situated within the hydrological catchment area of the groundwater, which was confirmed when identical strains of C. jejuni were isolated from groundwater and the dairy herd. In the present study, Campylobacter was not identified from groundwater, and dairy herds and other domestic animals in farms around the source were not investigated. However, results of bacterial, chemical and physical water analyses suggested strongly that a river contaminated by agricultural run-off from surrounding sheep and cattle grazing meadows was contaminating the source. In addition, the immediate area of the source was insufficiently protected and vulnerable to the environment. Wild birds, rodents, pigs and cows are all hosts for C. coli and could contaminate surface water easily [26,27,34,35]. Campylobacters are sensitive to many environmental factors and to chlorine [33,36,37], but a failure of the chlorination system could have resulted in the presence of a high concentration of C. coli. Waterborne outbreaks are often caused by several factors that act together [18,20,23,38]. In some outbreaks, heavy rainfall has been responsible for increasing agricultural run-off into rivers, although this did not occur before the outbreak in Gourdon. This underlies the need for increased frequency of routine water quality monitoring, chlorination of the water supply according to environmental conditions, and implementation of preventive measures. This outbreak raised the question of the delay in raising the alert. An outbreak can be recognised by: (i) the poor quality of water detected during regular controls; (ii) complaints about water quality; (iii) an incident in the supply and ℠or sewage water system during maintenance work; and (iv) an increased number of cases of gastroenteritis. In most instances, waterborne outbreaks are suspected when there are already a large number of declared cases [3,12,18,28,38,39]. In the present outbreak, the initial warning was given by a GP who notified officials 5–10 days after the assumed date of onset. This highlights the role of GPs in the detection of an unusual event, particularly during the summer, when an increase in acute gastroenteritis is less expected than during winter [40,41]. In addition, a routine water quality monitoring system able to detect contamination of drinking water before the occurrence of illness and complaints about the water quality should be used as an early warning source. The retrospective cohort study of Gourdon residents allowed the scale and severity of the outbreak to be quantified precisely and the hypothesis that tap water was the vehicle of the outbreak to be tested. When provided with a cohort sample representative of the study population, the number of people affected can be estimated. However, as the exposure of the population was high, a relatively large sample size was needed to test the association between the disease and the exposure. Because of misclassification of exposure (unrecognised consumption of tap water), the attributable risk is probably underestimated. In the absence of obvious alternative explanations, water contamination is the most likely common cause of an explosive outbreak of AGE occurring in a large population. Tillett et al. [42] proposed categorising levels of evidence for waterborne outbreaks. A descriptive study suggesting a water-related outbreak, together with the identification of the same pathogen in stools and water, should be sufficient to conclude that an outbreak is associated strongly with water consumption. Faecal indicators of water quality are currently insufficiently sensitive for some pathogens, notably viruses and parasites (e.g., Cryptosporidium). Therefore, in the absence of evidence from an analytical study to demonstrate an association, the link between illness and water can only be classified as probable or possible. However, total coliform counts are good indicators of faecal contamination with other pathogens (e.g., rotaviruses, E. coli, Campylobacter) [43]. In conclusion, the waterborne outbreak described in the present study was most probably caused by faecal contamination of source A by groundwater, resulting from deficiencies in the maintenance of the water distribution system. Corrective measures were implemented based on the results of the investigation. To our knowledge, this is the first documented waterborne outbreak involving human C. coli infection. In-depth environmental investigations to identify the origin of water contamination are important in order to better understand microbial transmission from the environment and to implement appropriate controls and preventive measures. In addition, surveys on vehicles and vectors that transmit Campylobacter between hosts are crucial in order to better understand the epidemiology of these organisms.

What caused the event?

290

A large multi-pathogen waterborne community outbreak linked to faecal contamination of a groundwater system, France, 2000

A large waterborne outbreak of infection that occurred during August 2000 in a local community in France was investigated initially via a rapid survey of visits to local physicians. A retrospective cohort study was then conducted on a random cluster sample of residents. Of 709 residents interviewed, 202 (28.5%) were definite cases (at least three liquid stools ℠day or vomiting) and 62 (8.7%) were probable cases (less than three liquid stools ℠day or abdominal pain). Those who had drunk tap water had a threefold increased risk for illness (95% CI 2.4–4.0). The risk increased with the amount of water consumed (chi-square trend: p < 0.0001). Bacteriological analyses of stools were performed for 35 patients and virological analyses for 24 patients. Campylobacter coli, group A rotavirus and norovirus were detected in 31.5%, 71.0% and 21% of samples, respectively. An extensive environmental investigation concluded that a groundwater source to this community had probably been contaminated by agricultural run-off, and a failure in the chlorination system was identified. This is the first documented waterborne outbreak of infection involving human C. coli infections. A better understanding of the factors influencing campylobacter transmission between hosts is required. Keywords Campylobacter coli, epidemiology, gastroenteritis, norovirus, rotavirus, waterborne outbreak INTRODUCTION Contaminated drinking water causes extensive outbreaks of illness because of the large number of people served by water supply facilities. Detection of such outbreaks requires the identification of an increase in illness (usually the rate of gastrointestinal disease) in the exposed population and confirmation that water was the route of transmission [1]. Many countries have routine reporting systems for detecting foodborne and waterborne outbreaks of infection. However, in routine practice it is not easy to detect an increase in clinical cases and to link this increase to waterborne transmission; thus waterborne outbreaks are often unrecognised and underestimated [2]. Microbiological examination of water is more complex than examination of stools, and there is often a failure to detect pathogens in water. Norovirus, Cryptosporidium parvum, Giardia intestinalis and Campylobacter jejuni are the pathogens identified most frequently [3]. The present study describes a large waterborne community outbreak of infection with multiple pathogens, including Campylobacter coli. METHODS AND MATERIALS Background On 23 August 2000, a general practitioner (GP) informed the local district health department that 16 cases of acute gastroenteritis (AGE) had occurred among residents of a holiday camp in the Gourdon community since 21 August. Gourdon, which is located in the south-west of France, has a population of 4888 inhabitants, which increases by 50% in the July– August tourist season. Investigations at the holiday camp concluded that foodborne transmission was very unlikely, but found that tap water was highly contaminated with faecal coliforms. Preliminary information collected from physicians in Gourdon indicated an increase of the number of consultations for AGE. Gourdon tap water is supplied by two underground water facilities: facility A serving 3800 water consumers in Gourdon; and facility B serving 1200 water consumers in Gourdon and the surrounding communities. Water treatment in facility A consists of chlorination, which involves two automatic pumps working in rotation. The water then passes through two semi in-ground reservoir tanks (1100 m3 and 200 m3, respectively) before distribution. Facility B water is mixed with a fraction of facility A water in another 500 m3 semi in-ground reservoir tank before being distributed. This mixing with the treated water from facility A is intended to disinfect the water from facility B. Since the increase in AGE was noted only in Gourdon, contaminated tap water from facility A was suspected to be the origin of the outbreak. Therefore, the population of Gourdon was informed on 25 August to avoid the consumption of tap water or to boil it for 5 min. Survey of gastroenteritis-related medical consultations To confirm the outbreak, the 11 GPs in Gourdon and the local hospital emergency department were requested to provide the number of total daily visits and the number of cases of gastroenteritis cases occurring between 1 August and 11 September 2000. Retrospective cohort study A retrospective cohort study was performed on a sample of Gourdon households between 5 and 15 September 2000. Domestic households were selected randomly from the telephone directory. A definite case was defined as a Gourdon resident who had at least three liquid stools ℠day or vomiting between 1 August and 3 September, and who had been present in Gourdon for ‡ 1 day between 1 and 31 August; if less than three liquid stools ℠day or abdominal pain had occurred, the case was classified as probable. Exposure was defined as the consumption of tap water before 25 August 2000. A standardised telephone questionnaire was completed to obtain demographic data (gender, date of birth), food consumption (commercially prepared dishes, sea-food, raw milk, pastry), tap water consumption (before and after 25 August, daily number of glasses of water, consumption of boiled tap water), symptoms of gastroenteritis (date of onset, number of liquid stools ℠day, vomiting, nausea, abdominal pain, fever), medical care (in- or out-patient) and consequences on daily activities (sick leave, bed confinement) for each member of the selected household. Under the hypothesis that 60% of the population consumed tap water, an estimated 550 people were required for inclusion in the study to identify a relative risk (RR) of ‡ 3, with a 5% a risk and a power of 80%. Based on an average of two individuals per household (INSERM 1999 population census), 300 households were required for inclusion in the survey. The association between AGE and the consumption of tap water was assessed by calculating the risk ratio (RR) and 95% CI, taking into account a household cluster design effect. A dose-effect relationship was evaluated by analysing the RR trend for different categories of daily number of glasses of tap water consumed (chi-square trend). Proportions were compared using the corrected chi-square test, and means were compared using the Student test or Mann and Witney test. Data were analysed with Epi-Info 6.04c (CDC, Atlanta, GA, USA). The gender and age distribution of the cohort population were compared with that of the general population of Gourdon (INSERM 1999 population census) to assess whether it was representative. Under the hypothesis of representation, the total number of cases of gastroenteritis was estimated by applying the attack rate to the total population of Gourdon, including the tourist population (data provided by the Gourdon tourism office). The most likely period for contamination of the water distribution system was based on the epidemic curve and the minimal incubation period of the pathogens involved. Microbiological analyses of stools Stool samples were examined for the presence of Salmonella spp., Shigella spp., Staphylococcus aureus, Campylobacter spp., Yersinia enterocolitica, Escherichia coli, enterococci, rotavirus groups A and C, astrovirus, calicivirus, adenovirus types 40 and 41, enterovirus, hepatitis A virus and Cryptosporidum. Viruses were detected using an enzyme immunoassay for group A rotavirus, astrovirus and adenovirus 40 ℠41, and by RT-PCR for calicivirus, group C rotavirus, enterovirus and hepatitis A virus [4–6]. Campylobacter isolates were sent to the National Reference Centre for Campylobacter for identification and fingerprinting. Species identification was by means of phenotypic tests (API Campy; bioMe´rieux, Lyon, France), with use of PCR to differentiate Campylobacter jejuni from Camplylobacter coli [7]. Random amplified polymorphic DNA fingerprinting was performed with primer 3881, 5¢-AACGCGCAAC. Environmental investigation An environmental and microbiological investigation of the water distribution system was performed. The inspection concerned the water catchments, the water distribution and sewage system, agricultural practices and private households around the sources. Bacteriological (Escherichia coli, enterococci, faecal streptococci, thermotolerant coliforms, sulphite-reducing clostridia) and chemical (chlorine) analyses were performed on tap water and on the water sampled from the two groundwater sources. After 23 August, the water chlorine concentration and the presence of coliform bacteria were monitored daily at several sites of the water system before and after the resort treatment. Examinations for viruses (group A rotavirus, astrovirus, norovirus, enterovirus, hepatitis A virus) and Campylobacter spp. were undertaken on two 10-mL tap water samples collected inhouse on 23 August, and on two 10-mL samples from groundwater sources A and B on 25 August. Viruses were detected by RT-PCR, followed by hybridisation with a specific primer to improve sensitivity, while C. jejuni and C. coli were detected by PCR with specific primers. Group A rotavirus isolates detected in stools and in water were sequenced using Beg 9 and End 9 primers [8] and compared using the Infobiogene GCG software (http://www.infobiogen.fr). Bacteriological (revived 36C, revived 22C, thermotolerant coliforms, streptococci and faecal enterococci), physical (temperature, conductivity) and chemical (oxygen consumption, ammonium, nitrates, phosphorus) parameters from source A and the river were analysed. Trends were compared for an 8-week period between 21 December 2000 and 7 February 2001. RESULTS Epidemiology Between 1 August and 11 September 2000, 7104 Gourdon residents consulted a GP or the local hospital emergency department. Between 24 and 28 August, gastroenteritis accounted for 44% (479 ℠1097) of all medical visits, compared with 6% (126 ℠2185) during the first 15 days of August (Fig. 1). Among the 498 households contacted for the household survey, 198 were excluded (103 did not respond, 49 did not fulfil the inclusion criteria, 43 refused to participate, and three for unknown reasons). The questionnaire was completed for 300 households (709 individuals). Data were missing concerning gender for one individual and age for 11 individuals. Overall, 331 (46.7%; 95% CI 44.5–49.0) of respondents were male and 377 (53.2%; 95% CI 50.9–55.5) were female. The median age was 44 years (range < 1–94 years). The age and gender distribution of the sample was similar to that of the general population. During the outbreak period, 264 of the 709 respondents (attack rate (AR) 37.2%) had been ill, with 202 definite cases (AR 28.5%; 95% CI 24.6– 32.4) and 62 probable cases (AR 8.7%; 95% CI 6.5– 11.0). The AR was higher in females (31.8%) than in males (24.8%), and was highest in children aged < 6 years (42.1%, Table 1). The epidemic curve (Fig. 2) suggests that the outbreak started between 14 and 19 August. The shape of the epidemic curve in the retrospective cohort study was very similar to that based on the number of visits to GPs for gastroenteritis. Patients complained mainly of diarrhoea, abdominal pain and nausea (Table 2). No cases of bloody diarrhoea were notified. The mean duration of symptoms was 4.5 days (median 3, range 1–30 days). Among definite cases, 52% had visited a GP, 42% had taken sick leave for ‡ 1 day, 32% had required bed rest, and 2.9% (six patients) had been hospitalised. In total, 336 (47.4%) respondents had drunk tap water before 25 August, whereas 646 (91.1%) did not drink tap water after 25 August. The risk for illness, taking into account the sample cluster design effect of 1.4, was three-fold greater (95% CI 2.2–4.1) for those who had drunk tap water before 25 August, and increased with the amount of water consumed (chi-square trend: p < 0.0001; Table 3). The risk of gastroenteritis associated with drinking tap water did not change substantially after adjustment for age (RRMH 3.3, 95% CI 2.5–4.4), and there was no modification effect according to age. Food consumption was not associated with illness. Applying the AR of 37% to the total population of Gourdon exposed to the water supply (4888 residents + 2200 tourists), the number of individuals affected by this outbreak was estimated to be 2600 (95% CI 2400–2900), assuming an exposure to tap water and a risk for gastroenteritis among the tourists similar to that among the inhabitants. Microbiological stool analyses Among the 35 stool samples analysed for bacteria, 11 (31.5%) were positive for C. coli. Twenty-four of the 35 samples were also tested for viruses: rotavirus A was detected in 17 (71.0%) and norovirus in five (21.0%) samples, while nine (37.5%) samples were co-infected (four C. coli + rotavirus; one C. coli + norovirus; three rotavirus + norovirus; and one C. coli + rotavirus + norovirus). Of the 11 C. coli isolates, six were characterised as biotype 2, and were identical following RAPD analysis, while the remaining five were of biotype 1. Rotavirus typing by RTPCR identified a single genotype P-(8), G1. Molecular characterisation of the norovirus isolates identified two different patterns, with three isolates being identical and close to the Saragota strain (genogroup I) and two close to the new GGIIb variant (genogroup II). Environmental investigation All of the water samples from source B were negative, regardless of the site of the distribution system tested. Analyses of the water originating from source A, sampled after leaving the water-works on 23 and 25 August, revealed high concentrations of E. coli (2263 / 100 mL), enterococci (415 / 100 mL), total coliforms (90 / 100 mL), faecal streptococci (136 / 100 mL) and sulphite-reducing clostridia (10 / 100 mL), and an absence of chlorination. Rotavirus A genotype G1 was detected in the crude water sampled from source A. The analyses for enterovirus, astrovirus, norovirus, hepatitis A virus and Campylobacter were negative. Source A was situated in a valley containing a river, below a hamlet of 15 dwellings, including farms breeding cattle, pork and sheep. The source was not supposed to be supplied by the river. The river resulted from the convergence of about ten different streams, with one originating close to a slaughterhouse, and was exposed to agricultural run-off and manure effluent; sheep and cattle grazed in meadows bordering the river, and a large amount of sewage from water treatment facilities was sprayed on to land close to the river. Cowsheds were close to the water catchment area. The immediate protected area around source A was too small and littered with building rubbish. Air vents without protection in the door made the source accessible to animals and insects. Bacteriological analyses of the water samples from source A and the river revealed that the contamination was of both human and animal origin, with a higher concentration in the river. The levels and trends of the chemical parameters of source A were similar to those of the river, strongly suggesting that the river was connected to source A (Fig. 3). A retrospective review of routine data collected between 1991 and 2000 showed irregularities in the maintenance of the water distribution system and the water treatment plant, particularly during the month of August. In the hamlet, the sewage system of the 15 households did not comply with sanitary rules. All but three ejected waste either into a cesspool or superficially without treatment. Three regions of gene 9 of the rotavirus detected in stools and water were compared. Sequencing showed that the strains were closely related but not identical. Assuming a likely onset of the outbreak between 14 and 19 August, and taking into account a minimum incubation period of 2 days for rotavirus A, norovirus and C. coli, the contamination of the water distribution system probably occurred between 12 and 17 August. Contamination probably persisted until 28 August, as intensive chlorination, beginning on 25 August, was effective only from 28 August onwards (Fig. 2). DISCUSSION The present descriptive and retrospective cohort studies confirmed that the outbreak of infection in the community of Gourdon was related to the consumption of tap water and affected 2600 individuals. The outbreak involved multiple pathogens, with group A rotavirus and C. coli being the most frequent, followed by norovirus. Results of environmental investigations suggested that water contamination was related to the contamination of the source, combined with a failure of the water chlorination system. The association with tap water consumption, the dose-effect relationship, and the fact that multiple pathogens were involved, were all strongly in favour of tap water as the origin of infection. Moreover, the number of cases decreased dramatically after the water restriction advice and the implementation of control measures on the water supply. No food products explored in the study were associated with illness. An AR of 14.5% and 31% was observed among adults and children aged < 6 years, respectively, who reported that they had not consumed tap water. This relatively high rate could be explained by several phenomena. Many individuals may have consumed tap water without being aware of it, e.g., in drinks or ice cubes consumed outside of the home, via raw fruits and vegetables, dishes prepared with unboiled water, or while cleaning their teeth. It can be hypothesised that people who became ill without being aware of having consumed tap water were exposed to small amounts of water and had been subjected to a small infective dose. In addition, children, more than adults, may have been contaminated through person-to-person transmission (e.g., family contact), which is the most common route of transmission for rotavirus and calicivirus. Waterborne contamination can affect large numbers of individuals [3,9–11]. In the present study, assuming a 50% increase in the population during the tourist season, the number of affected people was estimated to be between 2400 and 2900. The cohort study showed that almost half (46.5%) of the cases consulted a GP. Applying this proportion to the estimated number of cases yielded an estimated 1116–1349 ill individuals who consulted a GP for gastroenteritis, similar to the number of consultations for gastroenteritis actually reported by the GPs (n = 1037). The stool sample analyses revealed the presence of multiple pathogens (rotavirus A, norovirus genogroups I and II and C. coli) as single or co-infections. Except for C. coli, these pathogens have been implicated in numerous waterborne outbreaks, suggesting an influx of sewage into the water supply system [10,12–18]. Rotaviruses are known to be responsible for large epidemics during winter seasons, affecting particularly children aged < 3 years. These epidemics are caused by the faecal route of contamination (personto-person transmission). The high proportion of stools positive for rotavirus (71%) is uncommon in waterborne outbreaks and probably reflects massive contamination [15,16,19]. Different molecular sequences of rotavirus A genotype G1 found in stools and water indicated a human faecal source of infection. In the present study, contamination of water by sewage could be linked with the large amount of sewage from the water purification plant sprayed on to land close to the river and the water catchments. In this outbreak, C. coli, which was identified in 32% of stools, was the second most frequent pathogen. To our knowledge, human C. coli contamination has not been described previously as a causative agent in waterborne outbreaks, unlike C. jejuni [11,20–25], although C. coli has been isolated from water [26]. The main symptoms observed (diarrhoea, abdominal pain, nausea) and the mean duration of the disease (4.5 days) were compatible with a Campylobacter infection. No bloody diarrhoea was reported, although bloody diarrhoea is not constant in Campylobacter infections [27]. The high proportion of individuals who had to take sick leave, and the fact that six individuals were hospitalised, suggested a relative severity of symptoms, which is much more compatible with Campylobacter infection than with virus infections. Among the hospitalised patients, 60% were infected with C. coli. Among patients whose stools were analysed and whose age was known, cases infected only with rotavirus were younger (mean age 49 years; range, 5 months to 95 years) than cases infected only with Campylobacter (mean age 66 years; range 23–88 years) or co-infected with rotavirus plus Campylobacter (mean age 69 years; range 2–91 years). Moreover, the proportion (52%) of medical visits was unusually high compared with the proportion of visits observed during other waterborne outbreaks [13,16,28]. Campylobacter needs specific atmospheric conditions for growth and survival and its detection can be very difficult [29,30]. Because of a rapid physiological transformation to a viable, though non-cultivable, state, it has been suggested that conventional culture methods often fail [31]. Nevertheless, in the present study, the failure to detect C. coli in water, and the detection of noroviruses in stools, could be caused by the small (10 mL) water samples examined. Ha¨nninen et al. [30] proposed that the volume used for detection of suspected pathogens in drinking water should be at least 8–10 L. Groundwater is considered widely to be microbiologically clean, and is usually used for drinking without treatment. However, groundwater can be contaminated by the environment [26,32]. Stanley et al. [33] showed that Campylobacter can occur in groundwater, with hydrological evidence suggesting that the source of contamination in their study was a dairy farm situated within the hydrological catchment area of the groundwater, which was confirmed when identical strains of C. jejuni were isolated from groundwater and the dairy herd. In the present study, Campylobacter was not identified from groundwater, and dairy herds and other domestic animals in farms around the source were not investigated. However, results of bacterial, chemical and physical water analyses suggested strongly that a river contaminated by agricultural run-off from surrounding sheep and cattle grazing meadows was contaminating the source. In addition, the immediate area of the source was insufficiently protected and vulnerable to the environment. Wild birds, rodents, pigs and cows are all hosts for C. coli and could contaminate surface water easily [26,27,34,35]. Campylobacters are sensitive to many environmental factors and to chlorine [33,36,37], but a failure of the chlorination system could have resulted in the presence of a high concentration of C. coli. Waterborne outbreaks are often caused by several factors that act together [18,20,23,38]. In some outbreaks, heavy rainfall has been responsible for increasing agricultural run-off into rivers, although this did not occur before the outbreak in Gourdon. This underlies the need for increased frequency of routine water quality monitoring, chlorination of the water supply according to environmental conditions, and implementation of preventive measures. This outbreak raised the question of the delay in raising the alert. An outbreak can be recognised by: (i) the poor quality of water detected during regular controls; (ii) complaints about water quality; (iii) an incident in the supply and ℠or sewage water system during maintenance work; and (iv) an increased number of cases of gastroenteritis. In most instances, waterborne outbreaks are suspected when there are already a large number of declared cases [3,12,18,28,38,39]. In the present outbreak, the initial warning was given by a GP who notified officials 5–10 days after the assumed date of onset. This highlights the role of GPs in the detection of an unusual event, particularly during the summer, when an increase in acute gastroenteritis is less expected than during winter [40,41]. In addition, a routine water quality monitoring system able to detect contamination of drinking water before the occurrence of illness and complaints about the water quality should be used as an early warning source. The retrospective cohort study of Gourdon residents allowed the scale and severity of the outbreak to be quantified precisely and the hypothesis that tap water was the vehicle of the outbreak to be tested. When provided with a cohort sample representative of the study population, the number of people affected can be estimated. However, as the exposure of the population was high, a relatively large sample size was needed to test the association between the disease and the exposure. Because of misclassification of exposure (unrecognised consumption of tap water), the attributable risk is probably underestimated. In the absence of obvious alternative explanations, water contamination is the most likely common cause of an explosive outbreak of AGE occurring in a large population. Tillett et al. [42] proposed categorising levels of evidence for waterborne outbreaks. A descriptive study suggesting a water-related outbreak, together with the identification of the same pathogen in stools and water, should be sufficient to conclude that an outbreak is associated strongly with water consumption. Faecal indicators of water quality are currently insufficiently sensitive for some pathogens, notably viruses and parasites (e.g., Cryptosporidium). Therefore, in the absence of evidence from an analytical study to demonstrate an association, the link between illness and water can only be classified as probable or possible. However, total coliform counts are good indicators of faecal contamination with other pathogens (e.g., rotaviruses, E. coli, Campylobacter) [43]. In conclusion, the waterborne outbreak described in the present study was most probably caused by faecal contamination of source A by groundwater, resulting from deficiencies in the maintenance of the water distribution system. Corrective measures were implemented based on the results of the investigation. To our knowledge, this is the first documented waterborne outbreak involving human C. coli infection. In-depth environmental investigations to identify the origin of water contamination are important in order to better understand microbial transmission from the environment and to implement appropriate controls and preventive measures. In addition, surveys on vehicles and vectors that transmit Campylobacter between hosts are crucial in order to better understand the epidemiology of these organisms.

What caused the event?

291

A large multi-pathogen waterborne community outbreak linked to faecal contamination of a groundwater system, France, 2000

A large waterborne outbreak of infection that occurred during August 2000 in a local community in France was investigated initially via a rapid survey of visits to local physicians. A retrospective cohort study was then conducted on a random cluster sample of residents. Of 709 residents interviewed, 202 (28.5%) were definite cases (at least three liquid stools ℠day or vomiting) and 62 (8.7%) were probable cases (less than three liquid stools ℠day or abdominal pain). Those who had drunk tap water had a threefold increased risk for illness (95% CI 2.4–4.0). The risk increased with the amount of water consumed (chi-square trend: p < 0.0001). Bacteriological analyses of stools were performed for 35 patients and virological analyses for 24 patients. Campylobacter coli, group A rotavirus and norovirus were detected in 31.5%, 71.0% and 21% of samples, respectively. An extensive environmental investigation concluded that a groundwater source to this community had probably been contaminated by agricultural run-off, and a failure in the chlorination system was identified. This is the first documented waterborne outbreak of infection involving human C. coli infections. A better understanding of the factors influencing campylobacter transmission between hosts is required. Keywords Campylobacter coli, epidemiology, gastroenteritis, norovirus, rotavirus, waterborne outbreak INTRODUCTION Contaminated drinking water causes extensive outbreaks of illness because of the large number of people served by water supply facilities. Detection of such outbreaks requires the identification of an increase in illness (usually the rate of gastrointestinal disease) in the exposed population and confirmation that water was the route of transmission [1]. Many countries have routine reporting systems for detecting foodborne and waterborne outbreaks of infection. However, in routine practice it is not easy to detect an increase in clinical cases and to link this increase to waterborne transmission; thus waterborne outbreaks are often unrecognised and underestimated [2]. Microbiological examination of water is more complex than examination of stools, and there is often a failure to detect pathogens in water. Norovirus, Cryptosporidium parvum, Giardia intestinalis and Campylobacter jejuni are the pathogens identified most frequently [3]. The present study describes a large waterborne community outbreak of infection with multiple pathogens, including Campylobacter coli. METHODS AND MATERIALS Background On 23 August 2000, a general practitioner (GP) informed the local district health department that 16 cases of acute gastroenteritis (AGE) had occurred among residents of a holiday camp in the Gourdon community since 21 August. Gourdon, which is located in the south-west of France, has a population of 4888 inhabitants, which increases by 50% in the July– August tourist season. Investigations at the holiday camp concluded that foodborne transmission was very unlikely, but found that tap water was highly contaminated with faecal coliforms. Preliminary information collected from physicians in Gourdon indicated an increase of the number of consultations for AGE. Gourdon tap water is supplied by two underground water facilities: facility A serving 3800 water consumers in Gourdon; and facility B serving 1200 water consumers in Gourdon and the surrounding communities. Water treatment in facility A consists of chlorination, which involves two automatic pumps working in rotation. The water then passes through two semi in-ground reservoir tanks (1100 m3 and 200 m3, respectively) before distribution. Facility B water is mixed with a fraction of facility A water in another 500 m3 semi in-ground reservoir tank before being distributed. This mixing with the treated water from facility A is intended to disinfect the water from facility B. Since the increase in AGE was noted only in Gourdon, contaminated tap water from facility A was suspected to be the origin of the outbreak. Therefore, the population of Gourdon was informed on 25 August to avoid the consumption of tap water or to boil it for 5 min. Survey of gastroenteritis-related medical consultations To confirm the outbreak, the 11 GPs in Gourdon and the local hospital emergency department were requested to provide the number of total daily visits and the number of cases of gastroenteritis cases occurring between 1 August and 11 September 2000. Retrospective cohort study A retrospective cohort study was performed on a sample of Gourdon households between 5 and 15 September 2000. Domestic households were selected randomly from the telephone directory. A definite case was defined as a Gourdon resident who had at least three liquid stools ℠day or vomiting between 1 August and 3 September, and who had been present in Gourdon for ‡ 1 day between 1 and 31 August; if less than three liquid stools ℠day or abdominal pain had occurred, the case was classified as probable. Exposure was defined as the consumption of tap water before 25 August 2000. A standardised telephone questionnaire was completed to obtain demographic data (gender, date of birth), food consumption (commercially prepared dishes, sea-food, raw milk, pastry), tap water consumption (before and after 25 August, daily number of glasses of water, consumption of boiled tap water), symptoms of gastroenteritis (date of onset, number of liquid stools ℠day, vomiting, nausea, abdominal pain, fever), medical care (in- or out-patient) and consequences on daily activities (sick leave, bed confinement) for each member of the selected household. Under the hypothesis that 60% of the population consumed tap water, an estimated 550 people were required for inclusion in the study to identify a relative risk (RR) of ‡ 3, with a 5% a risk and a power of 80%. Based on an average of two individuals per household (INSERM 1999 population census), 300 households were required for inclusion in the survey. The association between AGE and the consumption of tap water was assessed by calculating the risk ratio (RR) and 95% CI, taking into account a household cluster design effect. A dose-effect relationship was evaluated by analysing the RR trend for different categories of daily number of glasses of tap water consumed (chi-square trend). Proportions were compared using the corrected chi-square test, and means were compared using the Student test or Mann and Witney test. Data were analysed with Epi-Info 6.04c (CDC, Atlanta, GA, USA). The gender and age distribution of the cohort population were compared with that of the general population of Gourdon (INSERM 1999 population census) to assess whether it was representative. Under the hypothesis of representation, the total number of cases of gastroenteritis was estimated by applying the attack rate to the total population of Gourdon, including the tourist population (data provided by the Gourdon tourism office). The most likely period for contamination of the water distribution system was based on the epidemic curve and the minimal incubation period of the pathogens involved. Microbiological analyses of stools Stool samples were examined for the presence of Salmonella spp., Shigella spp., Staphylococcus aureus, Campylobacter spp., Yersinia enterocolitica, Escherichia coli, enterococci, rotavirus groups A and C, astrovirus, calicivirus, adenovirus types 40 and 41, enterovirus, hepatitis A virus and Cryptosporidum. Viruses were detected using an enzyme immunoassay for group A rotavirus, astrovirus and adenovirus 40 ℠41, and by RT-PCR for calicivirus, group C rotavirus, enterovirus and hepatitis A virus [4–6]. Campylobacter isolates were sent to the National Reference Centre for Campylobacter for identification and fingerprinting. Species identification was by means of phenotypic tests (API Campy; bioMe´rieux, Lyon, France), with use of PCR to differentiate Campylobacter jejuni from Camplylobacter coli [7]. Random amplified polymorphic DNA fingerprinting was performed with primer 3881, 5¢-AACGCGCAAC. Environmental investigation An environmental and microbiological investigation of the water distribution system was performed. The inspection concerned the water catchments, the water distribution and sewage system, agricultural practices and private households around the sources. Bacteriological (Escherichia coli, enterococci, faecal streptococci, thermotolerant coliforms, sulphite-reducing clostridia) and chemical (chlorine) analyses were performed on tap water and on the water sampled from the two groundwater sources. After 23 August, the water chlorine concentration and the presence of coliform bacteria were monitored daily at several sites of the water system before and after the resort treatment. Examinations for viruses (group A rotavirus, astrovirus, norovirus, enterovirus, hepatitis A virus) and Campylobacter spp. were undertaken on two 10-mL tap water samples collected inhouse on 23 August, and on two 10-mL samples from groundwater sources A and B on 25 August. Viruses were detected by RT-PCR, followed by hybridisation with a specific primer to improve sensitivity, while C. jejuni and C. coli were detected by PCR with specific primers. Group A rotavirus isolates detected in stools and in water were sequenced using Beg 9 and End 9 primers [8] and compared using the Infobiogene GCG software (http://www.infobiogen.fr). Bacteriological (revived 36C, revived 22C, thermotolerant coliforms, streptococci and faecal enterococci), physical (temperature, conductivity) and chemical (oxygen consumption, ammonium, nitrates, phosphorus) parameters from source A and the river were analysed. Trends were compared for an 8-week period between 21 December 2000 and 7 February 2001. RESULTS Epidemiology Between 1 August and 11 September 2000, 7104 Gourdon residents consulted a GP or the local hospital emergency department. Between 24 and 28 August, gastroenteritis accounted for 44% (479 ℠1097) of all medical visits, compared with 6% (126 ℠2185) during the first 15 days of August (Fig. 1). Among the 498 households contacted for the household survey, 198 were excluded (103 did not respond, 49 did not fulfil the inclusion criteria, 43 refused to participate, and three for unknown reasons). The questionnaire was completed for 300 households (709 individuals). Data were missing concerning gender for one individual and age for 11 individuals. Overall, 331 (46.7%; 95% CI 44.5–49.0) of respondents were male and 377 (53.2%; 95% CI 50.9–55.5) were female. The median age was 44 years (range < 1–94 years). The age and gender distribution of the sample was similar to that of the general population. During the outbreak period, 264 of the 709 respondents (attack rate (AR) 37.2%) had been ill, with 202 definite cases (AR 28.5%; 95% CI 24.6– 32.4) and 62 probable cases (AR 8.7%; 95% CI 6.5– 11.0). The AR was higher in females (31.8%) than in males (24.8%), and was highest in children aged < 6 years (42.1%, Table 1). The epidemic curve (Fig. 2) suggests that the outbreak started between 14 and 19 August. The shape of the epidemic curve in the retrospective cohort study was very similar to that based on the number of visits to GPs for gastroenteritis. Patients complained mainly of diarrhoea, abdominal pain and nausea (Table 2). No cases of bloody diarrhoea were notified. The mean duration of symptoms was 4.5 days (median 3, range 1–30 days). Among definite cases, 52% had visited a GP, 42% had taken sick leave for ‡ 1 day, 32% had required bed rest, and 2.9% (six patients) had been hospitalised. In total, 336 (47.4%) respondents had drunk tap water before 25 August, whereas 646 (91.1%) did not drink tap water after 25 August. The risk for illness, taking into account the sample cluster design effect of 1.4, was three-fold greater (95% CI 2.2–4.1) for those who had drunk tap water before 25 August, and increased with the amount of water consumed (chi-square trend: p < 0.0001; Table 3). The risk of gastroenteritis associated with drinking tap water did not change substantially after adjustment for age (RRMH 3.3, 95% CI 2.5–4.4), and there was no modification effect according to age. Food consumption was not associated with illness. Applying the AR of 37% to the total population of Gourdon exposed to the water supply (4888 residents + 2200 tourists), the number of individuals affected by this outbreak was estimated to be 2600 (95% CI 2400–2900), assuming an exposure to tap water and a risk for gastroenteritis among the tourists similar to that among the inhabitants. Microbiological stool analyses Among the 35 stool samples analysed for bacteria, 11 (31.5%) were positive for C. coli. Twenty-four of the 35 samples were also tested for viruses: rotavirus A was detected in 17 (71.0%) and norovirus in five (21.0%) samples, while nine (37.5%) samples were co-infected (four C. coli + rotavirus; one C. coli + norovirus; three rotavirus + norovirus; and one C. coli + rotavirus + norovirus). Of the 11 C. coli isolates, six were characterised as biotype 2, and were identical following RAPD analysis, while the remaining five were of biotype 1. Rotavirus typing by RTPCR identified a single genotype P-(8), G1. Molecular characterisation of the norovirus isolates identified two different patterns, with three isolates being identical and close to the Saragota strain (genogroup I) and two close to the new GGIIb variant (genogroup II). Environmental investigation All of the water samples from source B were negative, regardless of the site of the distribution system tested. Analyses of the water originating from source A, sampled after leaving the water-works on 23 and 25 August, revealed high concentrations of E. coli (2263 / 100 mL), enterococci (415 / 100 mL), total coliforms (90 / 100 mL), faecal streptococci (136 / 100 mL) and sulphite-reducing clostridia (10 / 100 mL), and an absence of chlorination. Rotavirus A genotype G1 was detected in the crude water sampled from source A. The analyses for enterovirus, astrovirus, norovirus, hepatitis A virus and Campylobacter were negative. Source A was situated in a valley containing a river, below a hamlet of 15 dwellings, including farms breeding cattle, pork and sheep. The source was not supposed to be supplied by the river. The river resulted from the convergence of about ten different streams, with one originating close to a slaughterhouse, and was exposed to agricultural run-off and manure effluent; sheep and cattle grazed in meadows bordering the river, and a large amount of sewage from water treatment facilities was sprayed on to land close to the river. Cowsheds were close to the water catchment area. The immediate protected area around source A was too small and littered with building rubbish. Air vents without protection in the door made the source accessible to animals and insects. Bacteriological analyses of the water samples from source A and the river revealed that the contamination was of both human and animal origin, with a higher concentration in the river. The levels and trends of the chemical parameters of source A were similar to those of the river, strongly suggesting that the river was connected to source A (Fig. 3). A retrospective review of routine data collected between 1991 and 2000 showed irregularities in the maintenance of the water distribution system and the water treatment plant, particularly during the month of August. In the hamlet, the sewage system of the 15 households did not comply with sanitary rules. All but three ejected waste either into a cesspool or superficially without treatment. Three regions of gene 9 of the rotavirus detected in stools and water were compared. Sequencing showed that the strains were closely related but not identical. Assuming a likely onset of the outbreak between 14 and 19 August, and taking into account a minimum incubation period of 2 days for rotavirus A, norovirus and C. coli, the contamination of the water distribution system probably occurred between 12 and 17 August. Contamination probably persisted until 28 August, as intensive chlorination, beginning on 25 August, was effective only from 28 August onwards (Fig. 2). DISCUSSION The present descriptive and retrospective cohort studies confirmed that the outbreak of infection in the community of Gourdon was related to the consumption of tap water and affected 2600 individuals. The outbreak involved multiple pathogens, with group A rotavirus and C. coli being the most frequent, followed by norovirus. Results of environmental investigations suggested that water contamination was related to the contamination of the source, combined with a failure of the water chlorination system. The association with tap water consumption, the dose-effect relationship, and the fact that multiple pathogens were involved, were all strongly in favour of tap water as the origin of infection. Moreover, the number of cases decreased dramatically after the water restriction advice and the implementation of control measures on the water supply. No food products explored in the study were associated with illness. An AR of 14.5% and 31% was observed among adults and children aged < 6 years, respectively, who reported that they had not consumed tap water. This relatively high rate could be explained by several phenomena. Many individuals may have consumed tap water without being aware of it, e.g., in drinks or ice cubes consumed outside of the home, via raw fruits and vegetables, dishes prepared with unboiled water, or while cleaning their teeth. It can be hypothesised that people who became ill without being aware of having consumed tap water were exposed to small amounts of water and had been subjected to a small infective dose. In addition, children, more than adults, may have been contaminated through person-to-person transmission (e.g., family contact), which is the most common route of transmission for rotavirus and calicivirus. Waterborne contamination can affect large numbers of individuals [3,9–11]. In the present study, assuming a 50% increase in the population during the tourist season, the number of affected people was estimated to be between 2400 and 2900. The cohort study showed that almost half (46.5%) of the cases consulted a GP. Applying this proportion to the estimated number of cases yielded an estimated 1116–1349 ill individuals who consulted a GP for gastroenteritis, similar to the number of consultations for gastroenteritis actually reported by the GPs (n = 1037). The stool sample analyses revealed the presence of multiple pathogens (rotavirus A, norovirus genogroups I and II and C. coli) as single or co-infections. Except for C. coli, these pathogens have been implicated in numerous waterborne outbreaks, suggesting an influx of sewage into the water supply system [10,12–18]. Rotaviruses are known to be responsible for large epidemics during winter seasons, affecting particularly children aged < 3 years. These epidemics are caused by the faecal route of contamination (personto-person transmission). The high proportion of stools positive for rotavirus (71%) is uncommon in waterborne outbreaks and probably reflects massive contamination [15,16,19]. Different molecular sequences of rotavirus A genotype G1 found in stools and water indicated a human faecal source of infection. In the present study, contamination of water by sewage could be linked with the large amount of sewage from the water purification plant sprayed on to land close to the river and the water catchments. In this outbreak, C. coli, which was identified in 32% of stools, was the second most frequent pathogen. To our knowledge, human C. coli contamination has not been described previously as a causative agent in waterborne outbreaks, unlike C. jejuni [11,20–25], although C. coli has been isolated from water [26]. The main symptoms observed (diarrhoea, abdominal pain, nausea) and the mean duration of the disease (4.5 days) were compatible with a Campylobacter infection. No bloody diarrhoea was reported, although bloody diarrhoea is not constant in Campylobacter infections [27]. The high proportion of individuals who had to take sick leave, and the fact that six individuals were hospitalised, suggested a relative severity of symptoms, which is much more compatible with Campylobacter infection than with virus infections. Among the hospitalised patients, 60% were infected with C. coli. Among patients whose stools were analysed and whose age was known, cases infected only with rotavirus were younger (mean age 49 years; range, 5 months to 95 years) than cases infected only with Campylobacter (mean age 66 years; range 23–88 years) or co-infected with rotavirus plus Campylobacter (mean age 69 years; range 2–91 years). Moreover, the proportion (52%) of medical visits was unusually high compared with the proportion of visits observed during other waterborne outbreaks [13,16,28]. Campylobacter needs specific atmospheric conditions for growth and survival and its detection can be very difficult [29,30]. Because of a rapid physiological transformation to a viable, though non-cultivable, state, it has been suggested that conventional culture methods often fail [31]. Nevertheless, in the present study, the failure to detect C. coli in water, and the detection of noroviruses in stools, could be caused by the small (10 mL) water samples examined. Ha¨nninen et al. [30] proposed that the volume used for detection of suspected pathogens in drinking water should be at least 8–10 L. Groundwater is considered widely to be microbiologically clean, and is usually used for drinking without treatment. However, groundwater can be contaminated by the environment [26,32]. Stanley et al. [33] showed that Campylobacter can occur in groundwater, with hydrological evidence suggesting that the source of contamination in their study was a dairy farm situated within the hydrological catchment area of the groundwater, which was confirmed when identical strains of C. jejuni were isolated from groundwater and the dairy herd. In the present study, Campylobacter was not identified from groundwater, and dairy herds and other domestic animals in farms around the source were not investigated. However, results of bacterial, chemical and physical water analyses suggested strongly that a river contaminated by agricultural run-off from surrounding sheep and cattle grazing meadows was contaminating the source. In addition, the immediate area of the source was insufficiently protected and vulnerable to the environment. Wild birds, rodents, pigs and cows are all hosts for C. coli and could contaminate surface water easily [26,27,34,35]. Campylobacters are sensitive to many environmental factors and to chlorine [33,36,37], but a failure of the chlorination system could have resulted in the presence of a high concentration of C. coli. Waterborne outbreaks are often caused by several factors that act together [18,20,23,38]. In some outbreaks, heavy rainfall has been responsible for increasing agricultural run-off into rivers, although this did not occur before the outbreak in Gourdon. This underlies the need for increased frequency of routine water quality monitoring, chlorination of the water supply according to environmental conditions, and implementation of preventive measures. This outbreak raised the question of the delay in raising the alert. An outbreak can be recognised by: (i) the poor quality of water detected during regular controls; (ii) complaints about water quality; (iii) an incident in the supply and ℠or sewage water system during maintenance work; and (iv) an increased number of cases of gastroenteritis. In most instances, waterborne outbreaks are suspected when there are already a large number of declared cases [3,12,18,28,38,39]. In the present outbreak, the initial warning was given by a GP who notified officials 5–10 days after the assumed date of onset. This highlights the role of GPs in the detection of an unusual event, particularly during the summer, when an increase in acute gastroenteritis is less expected than during winter [40,41]. In addition, a routine water quality monitoring system able to detect contamination of drinking water before the occurrence of illness and complaints about the water quality should be used as an early warning source. The retrospective cohort study of Gourdon residents allowed the scale and severity of the outbreak to be quantified precisely and the hypothesis that tap water was the vehicle of the outbreak to be tested. When provided with a cohort sample representative of the study population, the number of people affected can be estimated. However, as the exposure of the population was high, a relatively large sample size was needed to test the association between the disease and the exposure. Because of misclassification of exposure (unrecognised consumption of tap water), the attributable risk is probably underestimated. In the absence of obvious alternative explanations, water contamination is the most likely common cause of an explosive outbreak of AGE occurring in a large population. Tillett et al. [42] proposed categorising levels of evidence for waterborne outbreaks. A descriptive study suggesting a water-related outbreak, together with the identification of the same pathogen in stools and water, should be sufficient to conclude that an outbreak is associated strongly with water consumption. Faecal indicators of water quality are currently insufficiently sensitive for some pathogens, notably viruses and parasites (e.g., Cryptosporidium). Therefore, in the absence of evidence from an analytical study to demonstrate an association, the link between illness and water can only be classified as probable or possible. However, total coliform counts are good indicators of faecal contamination with other pathogens (e.g., rotaviruses, E. coli, Campylobacter) [43]. In conclusion, the waterborne outbreak described in the present study was most probably caused by faecal contamination of source A by groundwater, resulting from deficiencies in the maintenance of the water distribution system. Corrective measures were implemented based on the results of the investigation. To our knowledge, this is the first documented waterborne outbreak involving human C. coli infection. In-depth environmental investigations to identify the origin of water contamination are important in order to better understand microbial transmission from the environment and to implement appropriate controls and preventive measures. In addition, surveys on vehicles and vectors that transmit Campylobacter between hosts are crucial in order to better understand the epidemiology of these organisms.

What was the cause of the event?

292

A large multi-pathogen waterborne community outbreak linked to faecal contamination of a groundwater system, France, 2000

A large waterborne outbreak of infection that occurred during August 2000 in a local community in France was investigated initially via a rapid survey of visits to local physicians. A retrospective cohort study was then conducted on a random cluster sample of residents. Of 709 residents interviewed, 202 (28.5%) were definite cases (at least three liquid stools ℠day or vomiting) and 62 (8.7%) were probable cases (less than three liquid stools ℠day or abdominal pain). Those who had drunk tap water had a threefold increased risk for illness (95% CI 2.4–4.0). The risk increased with the amount of water consumed (chi-square trend: p < 0.0001). Bacteriological analyses of stools were performed for 35 patients and virological analyses for 24 patients. Campylobacter coli, group A rotavirus and norovirus were detected in 31.5%, 71.0% and 21% of samples, respectively. An extensive environmental investigation concluded that a groundwater source to this community had probably been contaminated by agricultural run-off, and a failure in the chlorination system was identified. This is the first documented waterborne outbreak of infection involving human C. coli infections. A better understanding of the factors influencing campylobacter transmission between hosts is required. Keywords Campylobacter coli, epidemiology, gastroenteritis, norovirus, rotavirus, waterborne outbreak INTRODUCTION Contaminated drinking water causes extensive outbreaks of illness because of the large number of people served by water supply facilities. Detection of such outbreaks requires the identification of an increase in illness (usually the rate of gastrointestinal disease) in the exposed population and confirmation that water was the route of transmission [1]. Many countries have routine reporting systems for detecting foodborne and waterborne outbreaks of infection. However, in routine practice it is not easy to detect an increase in clinical cases and to link this increase to waterborne transmission; thus waterborne outbreaks are often unrecognised and underestimated [2]. Microbiological examination of water is more complex than examination of stools, and there is often a failure to detect pathogens in water. Norovirus, Cryptosporidium parvum, Giardia intestinalis and Campylobacter jejuni are the pathogens identified most frequently [3]. The present study describes a large waterborne community outbreak of infection with multiple pathogens, including Campylobacter coli. METHODS AND MATERIALS Background On 23 August 2000, a general practitioner (GP) informed the local district health department that 16 cases of acute gastroenteritis (AGE) had occurred among residents of a holiday camp in the Gourdon community since 21 August. Gourdon, which is located in the south-west of France, has a population of 4888 inhabitants, which increases by 50% in the July– August tourist season. Investigations at the holiday camp concluded that foodborne transmission was very unlikely, but found that tap water was highly contaminated with faecal coliforms. Preliminary information collected from physicians in Gourdon indicated an increase of the number of consultations for AGE. Gourdon tap water is supplied by two underground water facilities: facility A serving 3800 water consumers in Gourdon; and facility B serving 1200 water consumers in Gourdon and the surrounding communities. Water treatment in facility A consists of chlorination, which involves two automatic pumps working in rotation. The water then passes through two semi in-ground reservoir tanks (1100 m3 and 200 m3, respectively) before distribution. Facility B water is mixed with a fraction of facility A water in another 500 m3 semi in-ground reservoir tank before being distributed. This mixing with the treated water from facility A is intended to disinfect the water from facility B. Since the increase in AGE was noted only in Gourdon, contaminated tap water from facility A was suspected to be the origin of the outbreak. Therefore, the population of Gourdon was informed on 25 August to avoid the consumption of tap water or to boil it for 5 min. Survey of gastroenteritis-related medical consultations To confirm the outbreak, the 11 GPs in Gourdon and the local hospital emergency department were requested to provide the number of total daily visits and the number of cases of gastroenteritis cases occurring between 1 August and 11 September 2000. Retrospective cohort study A retrospective cohort study was performed on a sample of Gourdon households between 5 and 15 September 2000. Domestic households were selected randomly from the telephone directory. A definite case was defined as a Gourdon resident who had at least three liquid stools ℠day or vomiting between 1 August and 3 September, and who had been present in Gourdon for ‡ 1 day between 1 and 31 August; if less than three liquid stools ℠day or abdominal pain had occurred, the case was classified as probable. Exposure was defined as the consumption of tap water before 25 August 2000. A standardised telephone questionnaire was completed to obtain demographic data (gender, date of birth), food consumption (commercially prepared dishes, sea-food, raw milk, pastry), tap water consumption (before and after 25 August, daily number of glasses of water, consumption of boiled tap water), symptoms of gastroenteritis (date of onset, number of liquid stools ℠day, vomiting, nausea, abdominal pain, fever), medical care (in- or out-patient) and consequences on daily activities (sick leave, bed confinement) for each member of the selected household. Under the hypothesis that 60% of the population consumed tap water, an estimated 550 people were required for inclusion in the study to identify a relative risk (RR) of ‡ 3, with a 5% a risk and a power of 80%. Based on an average of two individuals per household (INSERM 1999 population census), 300 households were required for inclusion in the survey. The association between AGE and the consumption of tap water was assessed by calculating the risk ratio (RR) and 95% CI, taking into account a household cluster design effect. A dose-effect relationship was evaluated by analysing the RR trend for different categories of daily number of glasses of tap water consumed (chi-square trend). Proportions were compared using the corrected chi-square test, and means were compared using the Student test or Mann and Witney test. Data were analysed with Epi-Info 6.04c (CDC, Atlanta, GA, USA). The gender and age distribution of the cohort population were compared with that of the general population of Gourdon (INSERM 1999 population census) to assess whether it was representative. Under the hypothesis of representation, the total number of cases of gastroenteritis was estimated by applying the attack rate to the total population of Gourdon, including the tourist population (data provided by the Gourdon tourism office). The most likely period for contamination of the water distribution system was based on the epidemic curve and the minimal incubation period of the pathogens involved. Microbiological analyses of stools Stool samples were examined for the presence of Salmonella spp., Shigella spp., Staphylococcus aureus, Campylobacter spp., Yersinia enterocolitica, Escherichia coli, enterococci, rotavirus groups A and C, astrovirus, calicivirus, adenovirus types 40 and 41, enterovirus, hepatitis A virus and Cryptosporidum. Viruses were detected using an enzyme immunoassay for group A rotavirus, astrovirus and adenovirus 40 ℠41, and by RT-PCR for calicivirus, group C rotavirus, enterovirus and hepatitis A virus [4–6]. Campylobacter isolates were sent to the National Reference Centre for Campylobacter for identification and fingerprinting. Species identification was by means of phenotypic tests (API Campy; bioMe´rieux, Lyon, France), with use of PCR to differentiate Campylobacter jejuni from Camplylobacter coli [7]. Random amplified polymorphic DNA fingerprinting was performed with primer 3881, 5¢-AACGCGCAAC. Environmental investigation An environmental and microbiological investigation of the water distribution system was performed. The inspection concerned the water catchments, the water distribution and sewage system, agricultural practices and private households around the sources. Bacteriological (Escherichia coli, enterococci, faecal streptococci, thermotolerant coliforms, sulphite-reducing clostridia) and chemical (chlorine) analyses were performed on tap water and on the water sampled from the two groundwater sources. After 23 August, the water chlorine concentration and the presence of coliform bacteria were monitored daily at several sites of the water system before and after the resort treatment. Examinations for viruses (group A rotavirus, astrovirus, norovirus, enterovirus, hepatitis A virus) and Campylobacter spp. were undertaken on two 10-mL tap water samples collected inhouse on 23 August, and on two 10-mL samples from groundwater sources A and B on 25 August. Viruses were detected by RT-PCR, followed by hybridisation with a specific primer to improve sensitivity, while C. jejuni and C. coli were detected by PCR with specific primers. Group A rotavirus isolates detected in stools and in water were sequenced using Beg 9 and End 9 primers [8] and compared using the Infobiogene GCG software (http://www.infobiogen.fr). Bacteriological (revived 36C, revived 22C, thermotolerant coliforms, streptococci and faecal enterococci), physical (temperature, conductivity) and chemical (oxygen consumption, ammonium, nitrates, phosphorus) parameters from source A and the river were analysed. Trends were compared for an 8-week period between 21 December 2000 and 7 February 2001. RESULTS Epidemiology Between 1 August and 11 September 2000, 7104 Gourdon residents consulted a GP or the local hospital emergency department. Between 24 and 28 August, gastroenteritis accounted for 44% (479 ℠1097) of all medical visits, compared with 6% (126 ℠2185) during the first 15 days of August (Fig. 1). Among the 498 households contacted for the household survey, 198 were excluded (103 did not respond, 49 did not fulfil the inclusion criteria, 43 refused to participate, and three for unknown reasons). The questionnaire was completed for 300 households (709 individuals). Data were missing concerning gender for one individual and age for 11 individuals. Overall, 331 (46.7%; 95% CI 44.5–49.0) of respondents were male and 377 (53.2%; 95% CI 50.9–55.5) were female. The median age was 44 years (range < 1–94 years). The age and gender distribution of the sample was similar to that of the general population. During the outbreak period, 264 of the 709 respondents (attack rate (AR) 37.2%) had been ill, with 202 definite cases (AR 28.5%; 95% CI 24.6– 32.4) and 62 probable cases (AR 8.7%; 95% CI 6.5– 11.0). The AR was higher in females (31.8%) than in males (24.8%), and was highest in children aged < 6 years (42.1%, Table 1). The epidemic curve (Fig. 2) suggests that the outbreak started between 14 and 19 August. The shape of the epidemic curve in the retrospective cohort study was very similar to that based on the number of visits to GPs for gastroenteritis. Patients complained mainly of diarrhoea, abdominal pain and nausea (Table 2). No cases of bloody diarrhoea were notified. The mean duration of symptoms was 4.5 days (median 3, range 1–30 days). Among definite cases, 52% had visited a GP, 42% had taken sick leave for ‡ 1 day, 32% had required bed rest, and 2.9% (six patients) had been hospitalised. In total, 336 (47.4%) respondents had drunk tap water before 25 August, whereas 646 (91.1%) did not drink tap water after 25 August. The risk for illness, taking into account the sample cluster design effect of 1.4, was three-fold greater (95% CI 2.2–4.1) for those who had drunk tap water before 25 August, and increased with the amount of water consumed (chi-square trend: p < 0.0001; Table 3). The risk of gastroenteritis associated with drinking tap water did not change substantially after adjustment for age (RRMH 3.3, 95% CI 2.5–4.4), and there was no modification effect according to age. Food consumption was not associated with illness. Applying the AR of 37% to the total population of Gourdon exposed to the water supply (4888 residents + 2200 tourists), the number of individuals affected by this outbreak was estimated to be 2600 (95% CI 2400–2900), assuming an exposure to tap water and a risk for gastroenteritis among the tourists similar to that among the inhabitants. Microbiological stool analyses Among the 35 stool samples analysed for bacteria, 11 (31.5%) were positive for C. coli. Twenty-four of the 35 samples were also tested for viruses: rotavirus A was detected in 17 (71.0%) and norovirus in five (21.0%) samples, while nine (37.5%) samples were co-infected (four C. coli + rotavirus; one C. coli + norovirus; three rotavirus + norovirus; and one C. coli + rotavirus + norovirus). Of the 11 C. coli isolates, six were characterised as biotype 2, and were identical following RAPD analysis, while the remaining five were of biotype 1. Rotavirus typing by RTPCR identified a single genotype P-(8), G1. Molecular characterisation of the norovirus isolates identified two different patterns, with three isolates being identical and close to the Saragota strain (genogroup I) and two close to the new GGIIb variant (genogroup II). Environmental investigation All of the water samples from source B were negative, regardless of the site of the distribution system tested. Analyses of the water originating from source A, sampled after leaving the water-works on 23 and 25 August, revealed high concentrations of E. coli (2263 / 100 mL), enterococci (415 / 100 mL), total coliforms (90 / 100 mL), faecal streptococci (136 / 100 mL) and sulphite-reducing clostridia (10 / 100 mL), and an absence of chlorination. Rotavirus A genotype G1 was detected in the crude water sampled from source A. The analyses for enterovirus, astrovirus, norovirus, hepatitis A virus and Campylobacter were negative. Source A was situated in a valley containing a river, below a hamlet of 15 dwellings, including farms breeding cattle, pork and sheep. The source was not supposed to be supplied by the river. The river resulted from the convergence of about ten different streams, with one originating close to a slaughterhouse, and was exposed to agricultural run-off and manure effluent; sheep and cattle grazed in meadows bordering the river, and a large amount of sewage from water treatment facilities was sprayed on to land close to the river. Cowsheds were close to the water catchment area. The immediate protected area around source A was too small and littered with building rubbish. Air vents without protection in the door made the source accessible to animals and insects. Bacteriological analyses of the water samples from source A and the river revealed that the contamination was of both human and animal origin, with a higher concentration in the river. The levels and trends of the chemical parameters of source A were similar to those of the river, strongly suggesting that the river was connected to source A (Fig. 3). A retrospective review of routine data collected between 1991 and 2000 showed irregularities in the maintenance of the water distribution system and the water treatment plant, particularly during the month of August. In the hamlet, the sewage system of the 15 households did not comply with sanitary rules. All but three ejected waste either into a cesspool or superficially without treatment. Three regions of gene 9 of the rotavirus detected in stools and water were compared. Sequencing showed that the strains were closely related but not identical. Assuming a likely onset of the outbreak between 14 and 19 August, and taking into account a minimum incubation period of 2 days for rotavirus A, norovirus and C. coli, the contamination of the water distribution system probably occurred between 12 and 17 August. Contamination probably persisted until 28 August, as intensive chlorination, beginning on 25 August, was effective only from 28 August onwards (Fig. 2). DISCUSSION The present descriptive and retrospective cohort studies confirmed that the outbreak of infection in the community of Gourdon was related to the consumption of tap water and affected 2600 individuals. The outbreak involved multiple pathogens, with group A rotavirus and C. coli being the most frequent, followed by norovirus. Results of environmental investigations suggested that water contamination was related to the contamination of the source, combined with a failure of the water chlorination system. The association with tap water consumption, the dose-effect relationship, and the fact that multiple pathogens were involved, were all strongly in favour of tap water as the origin of infection. Moreover, the number of cases decreased dramatically after the water restriction advice and the implementation of control measures on the water supply. No food products explored in the study were associated with illness. An AR of 14.5% and 31% was observed among adults and children aged < 6 years, respectively, who reported that they had not consumed tap water. This relatively high rate could be explained by several phenomena. Many individuals may have consumed tap water without being aware of it, e.g., in drinks or ice cubes consumed outside of the home, via raw fruits and vegetables, dishes prepared with unboiled water, or while cleaning their teeth. It can be hypothesised that people who became ill without being aware of having consumed tap water were exposed to small amounts of water and had been subjected to a small infective dose. In addition, children, more than adults, may have been contaminated through person-to-person transmission (e.g., family contact), which is the most common route of transmission for rotavirus and calicivirus. Waterborne contamination can affect large numbers of individuals [3,9–11]. In the present study, assuming a 50% increase in the population during the tourist season, the number of affected people was estimated to be between 2400 and 2900. The cohort study showed that almost half (46.5%) of the cases consulted a GP. Applying this proportion to the estimated number of cases yielded an estimated 1116–1349 ill individuals who consulted a GP for gastroenteritis, similar to the number of consultations for gastroenteritis actually reported by the GPs (n = 1037). The stool sample analyses revealed the presence of multiple pathogens (rotavirus A, norovirus genogroups I and II and C. coli) as single or co-infections. Except for C. coli, these pathogens have been implicated in numerous waterborne outbreaks, suggesting an influx of sewage into the water supply system [10,12–18]. Rotaviruses are known to be responsible for large epidemics during winter seasons, affecting particularly children aged < 3 years. These epidemics are caused by the faecal route of contamination (personto-person transmission). The high proportion of stools positive for rotavirus (71%) is uncommon in waterborne outbreaks and probably reflects massive contamination [15,16,19]. Different molecular sequences of rotavirus A genotype G1 found in stools and water indicated a human faecal source of infection. In the present study, contamination of water by sewage could be linked with the large amount of sewage from the water purification plant sprayed on to land close to the river and the water catchments. In this outbreak, C. coli, which was identified in 32% of stools, was the second most frequent pathogen. To our knowledge, human C. coli contamination has not been described previously as a causative agent in waterborne outbreaks, unlike C. jejuni [11,20–25], although C. coli has been isolated from water [26]. The main symptoms observed (diarrhoea, abdominal pain, nausea) and the mean duration of the disease (4.5 days) were compatible with a Campylobacter infection. No bloody diarrhoea was reported, although bloody diarrhoea is not constant in Campylobacter infections [27]. The high proportion of individuals who had to take sick leave, and the fact that six individuals were hospitalised, suggested a relative severity of symptoms, which is much more compatible with Campylobacter infection than with virus infections. Among the hospitalised patients, 60% were infected with C. coli. Among patients whose stools were analysed and whose age was known, cases infected only with rotavirus were younger (mean age 49 years; range, 5 months to 95 years) than cases infected only with Campylobacter (mean age 66 years; range 23–88 years) or co-infected with rotavirus plus Campylobacter (mean age 69 years; range 2–91 years). Moreover, the proportion (52%) of medical visits was unusually high compared with the proportion of visits observed during other waterborne outbreaks [13,16,28]. Campylobacter needs specific atmospheric conditions for growth and survival and its detection can be very difficult [29,30]. Because of a rapid physiological transformation to a viable, though non-cultivable, state, it has been suggested that conventional culture methods often fail [31]. Nevertheless, in the present study, the failure to detect C. coli in water, and the detection of noroviruses in stools, could be caused by the small (10 mL) water samples examined. Ha¨nninen et al. [30] proposed that the volume used for detection of suspected pathogens in drinking water should be at least 8–10 L. Groundwater is considered widely to be microbiologically clean, and is usually used for drinking without treatment. However, groundwater can be contaminated by the environment [26,32]. Stanley et al. [33] showed that Campylobacter can occur in groundwater, with hydrological evidence suggesting that the source of contamination in their study was a dairy farm situated within the hydrological catchment area of the groundwater, which was confirmed when identical strains of C. jejuni were isolated from groundwater and the dairy herd. In the present study, Campylobacter was not identified from groundwater, and dairy herds and other domestic animals in farms around the source were not investigated. However, results of bacterial, chemical and physical water analyses suggested strongly that a river contaminated by agricultural run-off from surrounding sheep and cattle grazing meadows was contaminating the source. In addition, the immediate area of the source was insufficiently protected and vulnerable to the environment. Wild birds, rodents, pigs and cows are all hosts for C. coli and could contaminate surface water easily [26,27,34,35]. Campylobacters are sensitive to many environmental factors and to chlorine [33,36,37], but a failure of the chlorination system could have resulted in the presence of a high concentration of C. coli. Waterborne outbreaks are often caused by several factors that act together [18,20,23,38]. In some outbreaks, heavy rainfall has been responsible for increasing agricultural run-off into rivers, although this did not occur before the outbreak in Gourdon. This underlies the need for increased frequency of routine water quality monitoring, chlorination of the water supply according to environmental conditions, and implementation of preventive measures. This outbreak raised the question of the delay in raising the alert. An outbreak can be recognised by: (i) the poor quality of water detected during regular controls; (ii) complaints about water quality; (iii) an incident in the supply and ℠or sewage water system during maintenance work; and (iv) an increased number of cases of gastroenteritis. In most instances, waterborne outbreaks are suspected when there are already a large number of declared cases [3,12,18,28,38,39]. In the present outbreak, the initial warning was given by a GP who notified officials 5–10 days after the assumed date of onset. This highlights the role of GPs in the detection of an unusual event, particularly during the summer, when an increase in acute gastroenteritis is less expected than during winter [40,41]. In addition, a routine water quality monitoring system able to detect contamination of drinking water before the occurrence of illness and complaints about the water quality should be used as an early warning source. The retrospective cohort study of Gourdon residents allowed the scale and severity of the outbreak to be quantified precisely and the hypothesis that tap water was the vehicle of the outbreak to be tested. When provided with a cohort sample representative of the study population, the number of people affected can be estimated. However, as the exposure of the population was high, a relatively large sample size was needed to test the association between the disease and the exposure. Because of misclassification of exposure (unrecognised consumption of tap water), the attributable risk is probably underestimated. In the absence of obvious alternative explanations, water contamination is the most likely common cause of an explosive outbreak of AGE occurring in a large population. Tillett et al. [42] proposed categorising levels of evidence for waterborne outbreaks. A descriptive study suggesting a water-related outbreak, together with the identification of the same pathogen in stools and water, should be sufficient to conclude that an outbreak is associated strongly with water consumption. Faecal indicators of water quality are currently insufficiently sensitive for some pathogens, notably viruses and parasites (e.g., Cryptosporidium). Therefore, in the absence of evidence from an analytical study to demonstrate an association, the link between illness and water can only be classified as probable or possible. However, total coliform counts are good indicators of faecal contamination with other pathogens (e.g., rotaviruses, E. coli, Campylobacter) [43]. In conclusion, the waterborne outbreak described in the present study was most probably caused by faecal contamination of source A by groundwater, resulting from deficiencies in the maintenance of the water distribution system. Corrective measures were implemented based on the results of the investigation. To our knowledge, this is the first documented waterborne outbreak involving human C. coli infection. In-depth environmental investigations to identify the origin of water contamination are important in order to better understand microbial transmission from the environment and to implement appropriate controls and preventive measures. In addition, surveys on vehicles and vectors that transmit Campylobacter between hosts are crucial in order to better understand the epidemiology of these organisms.

What source started the event?

293

A large multi-pathogen waterborne community outbreak linked to faecal contamination of a groundwater system, France, 2000

A large waterborne outbreak of infection that occurred during August 2000 in a local community in France was investigated initially via a rapid survey of visits to local physicians. A retrospective cohort study was then conducted on a random cluster sample of residents. Of 709 residents interviewed, 202 (28.5%) were definite cases (at least three liquid stools ℠day or vomiting) and 62 (8.7%) were probable cases (less than three liquid stools ℠day or abdominal pain). Those who had drunk tap water had a threefold increased risk for illness (95% CI 2.4–4.0). The risk increased with the amount of water consumed (chi-square trend: p < 0.0001). Bacteriological analyses of stools were performed for 35 patients and virological analyses for 24 patients. Campylobacter coli, group A rotavirus and norovirus were detected in 31.5%, 71.0% and 21% of samples, respectively. An extensive environmental investigation concluded that a groundwater source to this community had probably been contaminated by agricultural run-off, and a failure in the chlorination system was identified. This is the first documented waterborne outbreak of infection involving human C. coli infections. A better understanding of the factors influencing campylobacter transmission between hosts is required. Keywords Campylobacter coli, epidemiology, gastroenteritis, norovirus, rotavirus, waterborne outbreak INTRODUCTION Contaminated drinking water causes extensive outbreaks of illness because of the large number of people served by water supply facilities. Detection of such outbreaks requires the identification of an increase in illness (usually the rate of gastrointestinal disease) in the exposed population and confirmation that water was the route of transmission [1]. Many countries have routine reporting systems for detecting foodborne and waterborne outbreaks of infection. However, in routine practice it is not easy to detect an increase in clinical cases and to link this increase to waterborne transmission; thus waterborne outbreaks are often unrecognised and underestimated [2]. Microbiological examination of water is more complex than examination of stools, and there is often a failure to detect pathogens in water. Norovirus, Cryptosporidium parvum, Giardia intestinalis and Campylobacter jejuni are the pathogens identified most frequently [3]. The present study describes a large waterborne community outbreak of infection with multiple pathogens, including Campylobacter coli. METHODS AND MATERIALS Background On 23 August 2000, a general practitioner (GP) informed the local district health department that 16 cases of acute gastroenteritis (AGE) had occurred among residents of a holiday camp in the Gourdon community since 21 August. Gourdon, which is located in the south-west of France, has a population of 4888 inhabitants, which increases by 50% in the July– August tourist season. Investigations at the holiday camp concluded that foodborne transmission was very unlikely, but found that tap water was highly contaminated with faecal coliforms. Preliminary information collected from physicians in Gourdon indicated an increase of the number of consultations for AGE. Gourdon tap water is supplied by two underground water facilities: facility A serving 3800 water consumers in Gourdon; and facility B serving 1200 water consumers in Gourdon and the surrounding communities. Water treatment in facility A consists of chlorination, which involves two automatic pumps working in rotation. The water then passes through two semi in-ground reservoir tanks (1100 m3 and 200 m3, respectively) before distribution. Facility B water is mixed with a fraction of facility A water in another 500 m3 semi in-ground reservoir tank before being distributed. This mixing with the treated water from facility A is intended to disinfect the water from facility B. Since the increase in AGE was noted only in Gourdon, contaminated tap water from facility A was suspected to be the origin of the outbreak. Therefore, the population of Gourdon was informed on 25 August to avoid the consumption of tap water or to boil it for 5 min. Survey of gastroenteritis-related medical consultations To confirm the outbreak, the 11 GPs in Gourdon and the local hospital emergency department were requested to provide the number of total daily visits and the number of cases of gastroenteritis cases occurring between 1 August and 11 September 2000. Retrospective cohort study A retrospective cohort study was performed on a sample of Gourdon households between 5 and 15 September 2000. Domestic households were selected randomly from the telephone directory. A definite case was defined as a Gourdon resident who had at least three liquid stools ℠day or vomiting between 1 August and 3 September, and who had been present in Gourdon for ‡ 1 day between 1 and 31 August; if less than three liquid stools ℠day or abdominal pain had occurred, the case was classified as probable. Exposure was defined as the consumption of tap water before 25 August 2000. A standardised telephone questionnaire was completed to obtain demographic data (gender, date of birth), food consumption (commercially prepared dishes, sea-food, raw milk, pastry), tap water consumption (before and after 25 August, daily number of glasses of water, consumption of boiled tap water), symptoms of gastroenteritis (date of onset, number of liquid stools ℠day, vomiting, nausea, abdominal pain, fever), medical care (in- or out-patient) and consequences on daily activities (sick leave, bed confinement) for each member of the selected household. Under the hypothesis that 60% of the population consumed tap water, an estimated 550 people were required for inclusion in the study to identify a relative risk (RR) of ‡ 3, with a 5% a risk and a power of 80%. Based on an average of two individuals per household (INSERM 1999 population census), 300 households were required for inclusion in the survey. The association between AGE and the consumption of tap water was assessed by calculating the risk ratio (RR) and 95% CI, taking into account a household cluster design effect. A dose-effect relationship was evaluated by analysing the RR trend for different categories of daily number of glasses of tap water consumed (chi-square trend). Proportions were compared using the corrected chi-square test, and means were compared using the Student test or Mann and Witney test. Data were analysed with Epi-Info 6.04c (CDC, Atlanta, GA, USA). The gender and age distribution of the cohort population were compared with that of the general population of Gourdon (INSERM 1999 population census) to assess whether it was representative. Under the hypothesis of representation, the total number of cases of gastroenteritis was estimated by applying the attack rate to the total population of Gourdon, including the tourist population (data provided by the Gourdon tourism office). The most likely period for contamination of the water distribution system was based on the epidemic curve and the minimal incubation period of the pathogens involved. Microbiological analyses of stools Stool samples were examined for the presence of Salmonella spp., Shigella spp., Staphylococcus aureus, Campylobacter spp., Yersinia enterocolitica, Escherichia coli, enterococci, rotavirus groups A and C, astrovirus, calicivirus, adenovirus types 40 and 41, enterovirus, hepatitis A virus and Cryptosporidum. Viruses were detected using an enzyme immunoassay for group A rotavirus, astrovirus and adenovirus 40 ℠41, and by RT-PCR for calicivirus, group C rotavirus, enterovirus and hepatitis A virus [4–6]. Campylobacter isolates were sent to the National Reference Centre for Campylobacter for identification and fingerprinting. Species identification was by means of phenotypic tests (API Campy; bioMe´rieux, Lyon, France), with use of PCR to differentiate Campylobacter jejuni from Camplylobacter coli [7]. Random amplified polymorphic DNA fingerprinting was performed with primer 3881, 5¢-AACGCGCAAC. Environmental investigation An environmental and microbiological investigation of the water distribution system was performed. The inspection concerned the water catchments, the water distribution and sewage system, agricultural practices and private households around the sources. Bacteriological (Escherichia coli, enterococci, faecal streptococci, thermotolerant coliforms, sulphite-reducing clostridia) and chemical (chlorine) analyses were performed on tap water and on the water sampled from the two groundwater sources. After 23 August, the water chlorine concentration and the presence of coliform bacteria were monitored daily at several sites of the water system before and after the resort treatment. Examinations for viruses (group A rotavirus, astrovirus, norovirus, enterovirus, hepatitis A virus) and Campylobacter spp. were undertaken on two 10-mL tap water samples collected inhouse on 23 August, and on two 10-mL samples from groundwater sources A and B on 25 August. Viruses were detected by RT-PCR, followed by hybridisation with a specific primer to improve sensitivity, while C. jejuni and C. coli were detected by PCR with specific primers. Group A rotavirus isolates detected in stools and in water were sequenced using Beg 9 and End 9 primers [8] and compared using the Infobiogene GCG software (http://www.infobiogen.fr). Bacteriological (revived 36C, revived 22C, thermotolerant coliforms, streptococci and faecal enterococci), physical (temperature, conductivity) and chemical (oxygen consumption, ammonium, nitrates, phosphorus) parameters from source A and the river were analysed. Trends were compared for an 8-week period between 21 December 2000 and 7 February 2001. RESULTS Epidemiology Between 1 August and 11 September 2000, 7104 Gourdon residents consulted a GP or the local hospital emergency department. Between 24 and 28 August, gastroenteritis accounted for 44% (479 ℠1097) of all medical visits, compared with 6% (126 ℠2185) during the first 15 days of August (Fig. 1). Among the 498 households contacted for the household survey, 198 were excluded (103 did not respond, 49 did not fulfil the inclusion criteria, 43 refused to participate, and three for unknown reasons). The questionnaire was completed for 300 households (709 individuals). Data were missing concerning gender for one individual and age for 11 individuals. Overall, 331 (46.7%; 95% CI 44.5–49.0) of respondents were male and 377 (53.2%; 95% CI 50.9–55.5) were female. The median age was 44 years (range < 1–94 years). The age and gender distribution of the sample was similar to that of the general population. During the outbreak period, 264 of the 709 respondents (attack rate (AR) 37.2%) had been ill, with 202 definite cases (AR 28.5%; 95% CI 24.6– 32.4) and 62 probable cases (AR 8.7%; 95% CI 6.5– 11.0). The AR was higher in females (31.8%) than in males (24.8%), and was highest in children aged < 6 years (42.1%, Table 1). The epidemic curve (Fig. 2) suggests that the outbreak started between 14 and 19 August. The shape of the epidemic curve in the retrospective cohort study was very similar to that based on the number of visits to GPs for gastroenteritis. Patients complained mainly of diarrhoea, abdominal pain and nausea (Table 2). No cases of bloody diarrhoea were notified. The mean duration of symptoms was 4.5 days (median 3, range 1–30 days). Among definite cases, 52% had visited a GP, 42% had taken sick leave for ‡ 1 day, 32% had required bed rest, and 2.9% (six patients) had been hospitalised. In total, 336 (47.4%) respondents had drunk tap water before 25 August, whereas 646 (91.1%) did not drink tap water after 25 August. The risk for illness, taking into account the sample cluster design effect of 1.4, was three-fold greater (95% CI 2.2–4.1) for those who had drunk tap water before 25 August, and increased with the amount of water consumed (chi-square trend: p < 0.0001; Table 3). The risk of gastroenteritis associated with drinking tap water did not change substantially after adjustment for age (RRMH 3.3, 95% CI 2.5–4.4), and there was no modification effect according to age. Food consumption was not associated with illness. Applying the AR of 37% to the total population of Gourdon exposed to the water supply (4888 residents + 2200 tourists), the number of individuals affected by this outbreak was estimated to be 2600 (95% CI 2400–2900), assuming an exposure to tap water and a risk for gastroenteritis among the tourists similar to that among the inhabitants. Microbiological stool analyses Among the 35 stool samples analysed for bacteria, 11 (31.5%) were positive for C. coli. Twenty-four of the 35 samples were also tested for viruses: rotavirus A was detected in 17 (71.0%) and norovirus in five (21.0%) samples, while nine (37.5%) samples were co-infected (four C. coli + rotavirus; one C. coli + norovirus; three rotavirus + norovirus; and one C. coli + rotavirus + norovirus). Of the 11 C. coli isolates, six were characterised as biotype 2, and were identical following RAPD analysis, while the remaining five were of biotype 1. Rotavirus typing by RTPCR identified a single genotype P-(8), G1. Molecular characterisation of the norovirus isolates identified two different patterns, with three isolates being identical and close to the Saragota strain (genogroup I) and two close to the new GGIIb variant (genogroup II). Environmental investigation All of the water samples from source B were negative, regardless of the site of the distribution system tested. Analyses of the water originating from source A, sampled after leaving the water-works on 23 and 25 August, revealed high concentrations of E. coli (2263 / 100 mL), enterococci (415 / 100 mL), total coliforms (90 / 100 mL), faecal streptococci (136 / 100 mL) and sulphite-reducing clostridia (10 / 100 mL), and an absence of chlorination. Rotavirus A genotype G1 was detected in the crude water sampled from source A. The analyses for enterovirus, astrovirus, norovirus, hepatitis A virus and Campylobacter were negative. Source A was situated in a valley containing a river, below a hamlet of 15 dwellings, including farms breeding cattle, pork and sheep. The source was not supposed to be supplied by the river. The river resulted from the convergence of about ten different streams, with one originating close to a slaughterhouse, and was exposed to agricultural run-off and manure effluent; sheep and cattle grazed in meadows bordering the river, and a large amount of sewage from water treatment facilities was sprayed on to land close to the river. Cowsheds were close to the water catchment area. The immediate protected area around source A was too small and littered with building rubbish. Air vents without protection in the door made the source accessible to animals and insects. Bacteriological analyses of the water samples from source A and the river revealed that the contamination was of both human and animal origin, with a higher concentration in the river. The levels and trends of the chemical parameters of source A were similar to those of the river, strongly suggesting that the river was connected to source A (Fig. 3). A retrospective review of routine data collected between 1991 and 2000 showed irregularities in the maintenance of the water distribution system and the water treatment plant, particularly during the month of August. In the hamlet, the sewage system of the 15 households did not comply with sanitary rules. All but three ejected waste either into a cesspool or superficially without treatment. Three regions of gene 9 of the rotavirus detected in stools and water were compared. Sequencing showed that the strains were closely related but not identical. Assuming a likely onset of the outbreak between 14 and 19 August, and taking into account a minimum incubation period of 2 days for rotavirus A, norovirus and C. coli, the contamination of the water distribution system probably occurred between 12 and 17 August. Contamination probably persisted until 28 August, as intensive chlorination, beginning on 25 August, was effective only from 28 August onwards (Fig. 2). DISCUSSION The present descriptive and retrospective cohort studies confirmed that the outbreak of infection in the community of Gourdon was related to the consumption of tap water and affected 2600 individuals. The outbreak involved multiple pathogens, with group A rotavirus and C. coli being the most frequent, followed by norovirus. Results of environmental investigations suggested that water contamination was related to the contamination of the source, combined with a failure of the water chlorination system. The association with tap water consumption, the dose-effect relationship, and the fact that multiple pathogens were involved, were all strongly in favour of tap water as the origin of infection. Moreover, the number of cases decreased dramatically after the water restriction advice and the implementation of control measures on the water supply. No food products explored in the study were associated with illness. An AR of 14.5% and 31% was observed among adults and children aged < 6 years, respectively, who reported that they had not consumed tap water. This relatively high rate could be explained by several phenomena. Many individuals may have consumed tap water without being aware of it, e.g., in drinks or ice cubes consumed outside of the home, via raw fruits and vegetables, dishes prepared with unboiled water, or while cleaning their teeth. It can be hypothesised that people who became ill without being aware of having consumed tap water were exposed to small amounts of water and had been subjected to a small infective dose. In addition, children, more than adults, may have been contaminated through person-to-person transmission (e.g., family contact), which is the most common route of transmission for rotavirus and calicivirus. Waterborne contamination can affect large numbers of individuals [3,9–11]. In the present study, assuming a 50% increase in the population during the tourist season, the number of affected people was estimated to be between 2400 and 2900. The cohort study showed that almost half (46.5%) of the cases consulted a GP. Applying this proportion to the estimated number of cases yielded an estimated 1116–1349 ill individuals who consulted a GP for gastroenteritis, similar to the number of consultations for gastroenteritis actually reported by the GPs (n = 1037). The stool sample analyses revealed the presence of multiple pathogens (rotavirus A, norovirus genogroups I and II and C. coli) as single or co-infections. Except for C. coli, these pathogens have been implicated in numerous waterborne outbreaks, suggesting an influx of sewage into the water supply system [10,12–18]. Rotaviruses are known to be responsible for large epidemics during winter seasons, affecting particularly children aged < 3 years. These epidemics are caused by the faecal route of contamination (personto-person transmission). The high proportion of stools positive for rotavirus (71%) is uncommon in waterborne outbreaks and probably reflects massive contamination [15,16,19]. Different molecular sequences of rotavirus A genotype G1 found in stools and water indicated a human faecal source of infection. In the present study, contamination of water by sewage could be linked with the large amount of sewage from the water purification plant sprayed on to land close to the river and the water catchments. In this outbreak, C. coli, which was identified in 32% of stools, was the second most frequent pathogen. To our knowledge, human C. coli contamination has not been described previously as a causative agent in waterborne outbreaks, unlike C. jejuni [11,20–25], although C. coli has been isolated from water [26]. The main symptoms observed (diarrhoea, abdominal pain, nausea) and the mean duration of the disease (4.5 days) were compatible with a Campylobacter infection. No bloody diarrhoea was reported, although bloody diarrhoea is not constant in Campylobacter infections [27]. The high proportion of individuals who had to take sick leave, and the fact that six individuals were hospitalised, suggested a relative severity of symptoms, which is much more compatible with Campylobacter infection than with virus infections. Among the hospitalised patients, 60% were infected with C. coli. Among patients whose stools were analysed and whose age was known, cases infected only with rotavirus were younger (mean age 49 years; range, 5 months to 95 years) than cases infected only with Campylobacter (mean age 66 years; range 23–88 years) or co-infected with rotavirus plus Campylobacter (mean age 69 years; range 2–91 years). Moreover, the proportion (52%) of medical visits was unusually high compared with the proportion of visits observed during other waterborne outbreaks [13,16,28]. Campylobacter needs specific atmospheric conditions for growth and survival and its detection can be very difficult [29,30]. Because of a rapid physiological transformation to a viable, though non-cultivable, state, it has been suggested that conventional culture methods often fail [31]. Nevertheless, in the present study, the failure to detect C. coli in water, and the detection of noroviruses in stools, could be caused by the small (10 mL) water samples examined. Ha¨nninen et al. [30] proposed that the volume used for detection of suspected pathogens in drinking water should be at least 8–10 L. Groundwater is considered widely to be microbiologically clean, and is usually used for drinking without treatment. However, groundwater can be contaminated by the environment [26,32]. Stanley et al. [33] showed that Campylobacter can occur in groundwater, with hydrological evidence suggesting that the source of contamination in their study was a dairy farm situated within the hydrological catchment area of the groundwater, which was confirmed when identical strains of C. jejuni were isolated from groundwater and the dairy herd. In the present study, Campylobacter was not identified from groundwater, and dairy herds and other domestic animals in farms around the source were not investigated. However, results of bacterial, chemical and physical water analyses suggested strongly that a river contaminated by agricultural run-off from surrounding sheep and cattle grazing meadows was contaminating the source. In addition, the immediate area of the source was insufficiently protected and vulnerable to the environment. Wild birds, rodents, pigs and cows are all hosts for C. coli and could contaminate surface water easily [26,27,34,35]. Campylobacters are sensitive to many environmental factors and to chlorine [33,36,37], but a failure of the chlorination system could have resulted in the presence of a high concentration of C. coli. Waterborne outbreaks are often caused by several factors that act together [18,20,23,38]. In some outbreaks, heavy rainfall has been responsible for increasing agricultural run-off into rivers, although this did not occur before the outbreak in Gourdon. This underlies the need for increased frequency of routine water quality monitoring, chlorination of the water supply according to environmental conditions, and implementation of preventive measures. This outbreak raised the question of the delay in raising the alert. An outbreak can be recognised by: (i) the poor quality of water detected during regular controls; (ii) complaints about water quality; (iii) an incident in the supply and ℠or sewage water system during maintenance work; and (iv) an increased number of cases of gastroenteritis. In most instances, waterborne outbreaks are suspected when there are already a large number of declared cases [3,12,18,28,38,39]. In the present outbreak, the initial warning was given by a GP who notified officials 5–10 days after the assumed date of onset. This highlights the role of GPs in the detection of an unusual event, particularly during the summer, when an increase in acute gastroenteritis is less expected than during winter [40,41]. In addition, a routine water quality monitoring system able to detect contamination of drinking water before the occurrence of illness and complaints about the water quality should be used as an early warning source. The retrospective cohort study of Gourdon residents allowed the scale and severity of the outbreak to be quantified precisely and the hypothesis that tap water was the vehicle of the outbreak to be tested. When provided with a cohort sample representative of the study population, the number of people affected can be estimated. However, as the exposure of the population was high, a relatively large sample size was needed to test the association between the disease and the exposure. Because of misclassification of exposure (unrecognised consumption of tap water), the attributable risk is probably underestimated. In the absence of obvious alternative explanations, water contamination is the most likely common cause of an explosive outbreak of AGE occurring in a large population. Tillett et al. [42] proposed categorising levels of evidence for waterborne outbreaks. A descriptive study suggesting a water-related outbreak, together with the identification of the same pathogen in stools and water, should be sufficient to conclude that an outbreak is associated strongly with water consumption. Faecal indicators of water quality are currently insufficiently sensitive for some pathogens, notably viruses and parasites (e.g., Cryptosporidium). Therefore, in the absence of evidence from an analytical study to demonstrate an association, the link between illness and water can only be classified as probable or possible. However, total coliform counts are good indicators of faecal contamination with other pathogens (e.g., rotaviruses, E. coli, Campylobacter) [43]. In conclusion, the waterborne outbreak described in the present study was most probably caused by faecal contamination of source A by groundwater, resulting from deficiencies in the maintenance of the water distribution system. Corrective measures were implemented based on the results of the investigation. To our knowledge, this is the first documented waterborne outbreak involving human C. coli infection. In-depth environmental investigations to identify the origin of water contamination are important in order to better understand microbial transmission from the environment and to implement appropriate controls and preventive measures. In addition, surveys on vehicles and vectors that transmit Campylobacter between hosts are crucial in order to better understand the epidemiology of these organisms.

How was the event first detected?

294

A large multi-pathogen waterborne community outbreak linked to faecal contamination of a groundwater system, France, 2000

A large waterborne outbreak of infection that occurred during August 2000 in a local community in France was investigated initially via a rapid survey of visits to local physicians. A retrospective cohort study was then conducted on a random cluster sample of residents. Of 709 residents interviewed, 202 (28.5%) were definite cases (at least three liquid stools ℠day or vomiting) and 62 (8.7%) were probable cases (less than three liquid stools ℠day or abdominal pain). Those who had drunk tap water had a threefold increased risk for illness (95% CI 2.4–4.0). The risk increased with the amount of water consumed (chi-square trend: p < 0.0001). Bacteriological analyses of stools were performed for 35 patients and virological analyses for 24 patients. Campylobacter coli, group A rotavirus and norovirus were detected in 31.5%, 71.0% and 21% of samples, respectively. An extensive environmental investigation concluded that a groundwater source to this community had probably been contaminated by agricultural run-off, and a failure in the chlorination system was identified. This is the first documented waterborne outbreak of infection involving human C. coli infections. A better understanding of the factors influencing campylobacter transmission between hosts is required. Keywords Campylobacter coli, epidemiology, gastroenteritis, norovirus, rotavirus, waterborne outbreak INTRODUCTION Contaminated drinking water causes extensive outbreaks of illness because of the large number of people served by water supply facilities. Detection of such outbreaks requires the identification of an increase in illness (usually the rate of gastrointestinal disease) in the exposed population and confirmation that water was the route of transmission [1]. Many countries have routine reporting systems for detecting foodborne and waterborne outbreaks of infection. However, in routine practice it is not easy to detect an increase in clinical cases and to link this increase to waterborne transmission; thus waterborne outbreaks are often unrecognised and underestimated [2]. Microbiological examination of water is more complex than examination of stools, and there is often a failure to detect pathogens in water. Norovirus, Cryptosporidium parvum, Giardia intestinalis and Campylobacter jejuni are the pathogens identified most frequently [3]. The present study describes a large waterborne community outbreak of infection with multiple pathogens, including Campylobacter coli. METHODS AND MATERIALS Background On 23 August 2000, a general practitioner (GP) informed the local district health department that 16 cases of acute gastroenteritis (AGE) had occurred among residents of a holiday camp in the Gourdon community since 21 August. Gourdon, which is located in the south-west of France, has a population of 4888 inhabitants, which increases by 50% in the July– August tourist season. Investigations at the holiday camp concluded that foodborne transmission was very unlikely, but found that tap water was highly contaminated with faecal coliforms. Preliminary information collected from physicians in Gourdon indicated an increase of the number of consultations for AGE. Gourdon tap water is supplied by two underground water facilities: facility A serving 3800 water consumers in Gourdon; and facility B serving 1200 water consumers in Gourdon and the surrounding communities. Water treatment in facility A consists of chlorination, which involves two automatic pumps working in rotation. The water then passes through two semi in-ground reservoir tanks (1100 m3 and 200 m3, respectively) before distribution. Facility B water is mixed with a fraction of facility A water in another 500 m3 semi in-ground reservoir tank before being distributed. This mixing with the treated water from facility A is intended to disinfect the water from facility B. Since the increase in AGE was noted only in Gourdon, contaminated tap water from facility A was suspected to be the origin of the outbreak. Therefore, the population of Gourdon was informed on 25 August to avoid the consumption of tap water or to boil it for 5 min. Survey of gastroenteritis-related medical consultations To confirm the outbreak, the 11 GPs in Gourdon and the local hospital emergency department were requested to provide the number of total daily visits and the number of cases of gastroenteritis cases occurring between 1 August and 11 September 2000. Retrospective cohort study A retrospective cohort study was performed on a sample of Gourdon households between 5 and 15 September 2000. Domestic households were selected randomly from the telephone directory. A definite case was defined as a Gourdon resident who had at least three liquid stools ℠day or vomiting between 1 August and 3 September, and who had been present in Gourdon for ‡ 1 day between 1 and 31 August; if less than three liquid stools ℠day or abdominal pain had occurred, the case was classified as probable. Exposure was defined as the consumption of tap water before 25 August 2000. A standardised telephone questionnaire was completed to obtain demographic data (gender, date of birth), food consumption (commercially prepared dishes, sea-food, raw milk, pastry), tap water consumption (before and after 25 August, daily number of glasses of water, consumption of boiled tap water), symptoms of gastroenteritis (date of onset, number of liquid stools ℠day, vomiting, nausea, abdominal pain, fever), medical care (in- or out-patient) and consequences on daily activities (sick leave, bed confinement) for each member of the selected household. Under the hypothesis that 60% of the population consumed tap water, an estimated 550 people were required for inclusion in the study to identify a relative risk (RR) of ‡ 3, with a 5% a risk and a power of 80%. Based on an average of two individuals per household (INSERM 1999 population census), 300 households were required for inclusion in the survey. The association between AGE and the consumption of tap water was assessed by calculating the risk ratio (RR) and 95% CI, taking into account a household cluster design effect. A dose-effect relationship was evaluated by analysing the RR trend for different categories of daily number of glasses of tap water consumed (chi-square trend). Proportions were compared using the corrected chi-square test, and means were compared using the Student test or Mann and Witney test. Data were analysed with Epi-Info 6.04c (CDC, Atlanta, GA, USA). The gender and age distribution of the cohort population were compared with that of the general population of Gourdon (INSERM 1999 population census) to assess whether it was representative. Under the hypothesis of representation, the total number of cases of gastroenteritis was estimated by applying the attack rate to the total population of Gourdon, including the tourist population (data provided by the Gourdon tourism office). The most likely period for contamination of the water distribution system was based on the epidemic curve and the minimal incubation period of the pathogens involved. Microbiological analyses of stools Stool samples were examined for the presence of Salmonella spp., Shigella spp., Staphylococcus aureus, Campylobacter spp., Yersinia enterocolitica, Escherichia coli, enterococci, rotavirus groups A and C, astrovirus, calicivirus, adenovirus types 40 and 41, enterovirus, hepatitis A virus and Cryptosporidum. Viruses were detected using an enzyme immunoassay for group A rotavirus, astrovirus and adenovirus 40 ℠41, and by RT-PCR for calicivirus, group C rotavirus, enterovirus and hepatitis A virus [4–6]. Campylobacter isolates were sent to the National Reference Centre for Campylobacter for identification and fingerprinting. Species identification was by means of phenotypic tests (API Campy; bioMe´rieux, Lyon, France), with use of PCR to differentiate Campylobacter jejuni from Camplylobacter coli [7]. Random amplified polymorphic DNA fingerprinting was performed with primer 3881, 5¢-AACGCGCAAC. Environmental investigation An environmental and microbiological investigation of the water distribution system was performed. The inspection concerned the water catchments, the water distribution and sewage system, agricultural practices and private households around the sources. Bacteriological (Escherichia coli, enterococci, faecal streptococci, thermotolerant coliforms, sulphite-reducing clostridia) and chemical (chlorine) analyses were performed on tap water and on the water sampled from the two groundwater sources. After 23 August, the water chlorine concentration and the presence of coliform bacteria were monitored daily at several sites of the water system before and after the resort treatment. Examinations for viruses (group A rotavirus, astrovirus, norovirus, enterovirus, hepatitis A virus) and Campylobacter spp. were undertaken on two 10-mL tap water samples collected inhouse on 23 August, and on two 10-mL samples from groundwater sources A and B on 25 August. Viruses were detected by RT-PCR, followed by hybridisation with a specific primer to improve sensitivity, while C. jejuni and C. coli were detected by PCR with specific primers. Group A rotavirus isolates detected in stools and in water were sequenced using Beg 9 and End 9 primers [8] and compared using the Infobiogene GCG software (http://www.infobiogen.fr). Bacteriological (revived 36C, revived 22C, thermotolerant coliforms, streptococci and faecal enterococci), physical (temperature, conductivity) and chemical (oxygen consumption, ammonium, nitrates, phosphorus) parameters from source A and the river were analysed. Trends were compared for an 8-week period between 21 December 2000 and 7 February 2001. RESULTS Epidemiology Between 1 August and 11 September 2000, 7104 Gourdon residents consulted a GP or the local hospital emergency department. Between 24 and 28 August, gastroenteritis accounted for 44% (479 ℠1097) of all medical visits, compared with 6% (126 ℠2185) during the first 15 days of August (Fig. 1). Among the 498 households contacted for the household survey, 198 were excluded (103 did not respond, 49 did not fulfil the inclusion criteria, 43 refused to participate, and three for unknown reasons). The questionnaire was completed for 300 households (709 individuals). Data were missing concerning gender for one individual and age for 11 individuals. Overall, 331 (46.7%; 95% CI 44.5–49.0) of respondents were male and 377 (53.2%; 95% CI 50.9–55.5) were female. The median age was 44 years (range < 1–94 years). The age and gender distribution of the sample was similar to that of the general population. During the outbreak period, 264 of the 709 respondents (attack rate (AR) 37.2%) had been ill, with 202 definite cases (AR 28.5%; 95% CI 24.6– 32.4) and 62 probable cases (AR 8.7%; 95% CI 6.5– 11.0). The AR was higher in females (31.8%) than in males (24.8%), and was highest in children aged < 6 years (42.1%, Table 1). The epidemic curve (Fig. 2) suggests that the outbreak started between 14 and 19 August. The shape of the epidemic curve in the retrospective cohort study was very similar to that based on the number of visits to GPs for gastroenteritis. Patients complained mainly of diarrhoea, abdominal pain and nausea (Table 2). No cases of bloody diarrhoea were notified. The mean duration of symptoms was 4.5 days (median 3, range 1–30 days). Among definite cases, 52% had visited a GP, 42% had taken sick leave for ‡ 1 day, 32% had required bed rest, and 2.9% (six patients) had been hospitalised. In total, 336 (47.4%) respondents had drunk tap water before 25 August, whereas 646 (91.1%) did not drink tap water after 25 August. The risk for illness, taking into account the sample cluster design effect of 1.4, was three-fold greater (95% CI 2.2–4.1) for those who had drunk tap water before 25 August, and increased with the amount of water consumed (chi-square trend: p < 0.0001; Table 3). The risk of gastroenteritis associated with drinking tap water did not change substantially after adjustment for age (RRMH 3.3, 95% CI 2.5–4.4), and there was no modification effect according to age. Food consumption was not associated with illness. Applying the AR of 37% to the total population of Gourdon exposed to the water supply (4888 residents + 2200 tourists), the number of individuals affected by this outbreak was estimated to be 2600 (95% CI 2400–2900), assuming an exposure to tap water and a risk for gastroenteritis among the tourists similar to that among the inhabitants. Microbiological stool analyses Among the 35 stool samples analysed for bacteria, 11 (31.5%) were positive for C. coli. Twenty-four of the 35 samples were also tested for viruses: rotavirus A was detected in 17 (71.0%) and norovirus in five (21.0%) samples, while nine (37.5%) samples were co-infected (four C. coli + rotavirus; one C. coli + norovirus; three rotavirus + norovirus; and one C. coli + rotavirus + norovirus). Of the 11 C. coli isolates, six were characterised as biotype 2, and were identical following RAPD analysis, while the remaining five were of biotype 1. Rotavirus typing by RTPCR identified a single genotype P-(8), G1. Molecular characterisation of the norovirus isolates identified two different patterns, with three isolates being identical and close to the Saragota strain (genogroup I) and two close to the new GGIIb variant (genogroup II). Environmental investigation All of the water samples from source B were negative, regardless of the site of the distribution system tested. Analyses of the water originating from source A, sampled after leaving the water-works on 23 and 25 August, revealed high concentrations of E. coli (2263 / 100 mL), enterococci (415 / 100 mL), total coliforms (90 / 100 mL), faecal streptococci (136 / 100 mL) and sulphite-reducing clostridia (10 / 100 mL), and an absence of chlorination. Rotavirus A genotype G1 was detected in the crude water sampled from source A. The analyses for enterovirus, astrovirus, norovirus, hepatitis A virus and Campylobacter were negative. Source A was situated in a valley containing a river, below a hamlet of 15 dwellings, including farms breeding cattle, pork and sheep. The source was not supposed to be supplied by the river. The river resulted from the convergence of about ten different streams, with one originating close to a slaughterhouse, and was exposed to agricultural run-off and manure effluent; sheep and cattle grazed in meadows bordering the river, and a large amount of sewage from water treatment facilities was sprayed on to land close to the river. Cowsheds were close to the water catchment area. The immediate protected area around source A was too small and littered with building rubbish. Air vents without protection in the door made the source accessible to animals and insects. Bacteriological analyses of the water samples from source A and the river revealed that the contamination was of both human and animal origin, with a higher concentration in the river. The levels and trends of the chemical parameters of source A were similar to those of the river, strongly suggesting that the river was connected to source A (Fig. 3). A retrospective review of routine data collected between 1991 and 2000 showed irregularities in the maintenance of the water distribution system and the water treatment plant, particularly during the month of August. In the hamlet, the sewage system of the 15 households did not comply with sanitary rules. All but three ejected waste either into a cesspool or superficially without treatment. Three regions of gene 9 of the rotavirus detected in stools and water were compared. Sequencing showed that the strains were closely related but not identical. Assuming a likely onset of the outbreak between 14 and 19 August, and taking into account a minimum incubation period of 2 days for rotavirus A, norovirus and C. coli, the contamination of the water distribution system probably occurred between 12 and 17 August. Contamination probably persisted until 28 August, as intensive chlorination, beginning on 25 August, was effective only from 28 August onwards (Fig. 2). DISCUSSION The present descriptive and retrospective cohort studies confirmed that the outbreak of infection in the community of Gourdon was related to the consumption of tap water and affected 2600 individuals. The outbreak involved multiple pathogens, with group A rotavirus and C. coli being the most frequent, followed by norovirus. Results of environmental investigations suggested that water contamination was related to the contamination of the source, combined with a failure of the water chlorination system. The association with tap water consumption, the dose-effect relationship, and the fact that multiple pathogens were involved, were all strongly in favour of tap water as the origin of infection. Moreover, the number of cases decreased dramatically after the water restriction advice and the implementation of control measures on the water supply. No food products explored in the study were associated with illness. An AR of 14.5% and 31% was observed among adults and children aged < 6 years, respectively, who reported that they had not consumed tap water. This relatively high rate could be explained by several phenomena. Many individuals may have consumed tap water without being aware of it, e.g., in drinks or ice cubes consumed outside of the home, via raw fruits and vegetables, dishes prepared with unboiled water, or while cleaning their teeth. It can be hypothesised that people who became ill without being aware of having consumed tap water were exposed to small amounts of water and had been subjected to a small infective dose. In addition, children, more than adults, may have been contaminated through person-to-person transmission (e.g., family contact), which is the most common route of transmission for rotavirus and calicivirus. Waterborne contamination can affect large numbers of individuals [3,9–11]. In the present study, assuming a 50% increase in the population during the tourist season, the number of affected people was estimated to be between 2400 and 2900. The cohort study showed that almost half (46.5%) of the cases consulted a GP. Applying this proportion to the estimated number of cases yielded an estimated 1116–1349 ill individuals who consulted a GP for gastroenteritis, similar to the number of consultations for gastroenteritis actually reported by the GPs (n = 1037). The stool sample analyses revealed the presence of multiple pathogens (rotavirus A, norovirus genogroups I and II and C. coli) as single or co-infections. Except for C. coli, these pathogens have been implicated in numerous waterborne outbreaks, suggesting an influx of sewage into the water supply system [10,12–18]. Rotaviruses are known to be responsible for large epidemics during winter seasons, affecting particularly children aged < 3 years. These epidemics are caused by the faecal route of contamination (personto-person transmission). The high proportion of stools positive for rotavirus (71%) is uncommon in waterborne outbreaks and probably reflects massive contamination [15,16,19]. Different molecular sequences of rotavirus A genotype G1 found in stools and water indicated a human faecal source of infection. In the present study, contamination of water by sewage could be linked with the large amount of sewage from the water purification plant sprayed on to land close to the river and the water catchments. In this outbreak, C. coli, which was identified in 32% of stools, was the second most frequent pathogen. To our knowledge, human C. coli contamination has not been described previously as a causative agent in waterborne outbreaks, unlike C. jejuni [11,20–25], although C. coli has been isolated from water [26]. The main symptoms observed (diarrhoea, abdominal pain, nausea) and the mean duration of the disease (4.5 days) were compatible with a Campylobacter infection. No bloody diarrhoea was reported, although bloody diarrhoea is not constant in Campylobacter infections [27]. The high proportion of individuals who had to take sick leave, and the fact that six individuals were hospitalised, suggested a relative severity of symptoms, which is much more compatible with Campylobacter infection than with virus infections. Among the hospitalised patients, 60% were infected with C. coli. Among patients whose stools were analysed and whose age was known, cases infected only with rotavirus were younger (mean age 49 years; range, 5 months to 95 years) than cases infected only with Campylobacter (mean age 66 years; range 23–88 years) or co-infected with rotavirus plus Campylobacter (mean age 69 years; range 2–91 years). Moreover, the proportion (52%) of medical visits was unusually high compared with the proportion of visits observed during other waterborne outbreaks [13,16,28]. Campylobacter needs specific atmospheric conditions for growth and survival and its detection can be very difficult [29,30]. Because of a rapid physiological transformation to a viable, though non-cultivable, state, it has been suggested that conventional culture methods often fail [31]. Nevertheless, in the present study, the failure to detect C. coli in water, and the detection of noroviruses in stools, could be caused by the small (10 mL) water samples examined. Ha¨nninen et al. [30] proposed that the volume used for detection of suspected pathogens in drinking water should be at least 8–10 L. Groundwater is considered widely to be microbiologically clean, and is usually used for drinking without treatment. However, groundwater can be contaminated by the environment [26,32]. Stanley et al. [33] showed that Campylobacter can occur in groundwater, with hydrological evidence suggesting that the source of contamination in their study was a dairy farm situated within the hydrological catchment area of the groundwater, which was confirmed when identical strains of C. jejuni were isolated from groundwater and the dairy herd. In the present study, Campylobacter was not identified from groundwater, and dairy herds and other domestic animals in farms around the source were not investigated. However, results of bacterial, chemical and physical water analyses suggested strongly that a river contaminated by agricultural run-off from surrounding sheep and cattle grazing meadows was contaminating the source. In addition, the immediate area of the source was insufficiently protected and vulnerable to the environment. Wild birds, rodents, pigs and cows are all hosts for C. coli and could contaminate surface water easily [26,27,34,35]. Campylobacters are sensitive to many environmental factors and to chlorine [33,36,37], but a failure of the chlorination system could have resulted in the presence of a high concentration of C. coli. Waterborne outbreaks are often caused by several factors that act together [18,20,23,38]. In some outbreaks, heavy rainfall has been responsible for increasing agricultural run-off into rivers, although this did not occur before the outbreak in Gourdon. This underlies the need for increased frequency of routine water quality monitoring, chlorination of the water supply according to environmental conditions, and implementation of preventive measures. This outbreak raised the question of the delay in raising the alert. An outbreak can be recognised by: (i) the poor quality of water detected during regular controls; (ii) complaints about water quality; (iii) an incident in the supply and ℠or sewage water system during maintenance work; and (iv) an increased number of cases of gastroenteritis. In most instances, waterborne outbreaks are suspected when there are already a large number of declared cases [3,12,18,28,38,39]. In the present outbreak, the initial warning was given by a GP who notified officials 5–10 days after the assumed date of onset. This highlights the role of GPs in the detection of an unusual event, particularly during the summer, when an increase in acute gastroenteritis is less expected than during winter [40,41]. In addition, a routine water quality monitoring system able to detect contamination of drinking water before the occurrence of illness and complaints about the water quality should be used as an early warning source. The retrospective cohort study of Gourdon residents allowed the scale and severity of the outbreak to be quantified precisely and the hypothesis that tap water was the vehicle of the outbreak to be tested. When provided with a cohort sample representative of the study population, the number of people affected can be estimated. However, as the exposure of the population was high, a relatively large sample size was needed to test the association between the disease and the exposure. Because of misclassification of exposure (unrecognised consumption of tap water), the attributable risk is probably underestimated. In the absence of obvious alternative explanations, water contamination is the most likely common cause of an explosive outbreak of AGE occurring in a large population. Tillett et al. [42] proposed categorising levels of evidence for waterborne outbreaks. A descriptive study suggesting a water-related outbreak, together with the identification of the same pathogen in stools and water, should be sufficient to conclude that an outbreak is associated strongly with water consumption. Faecal indicators of water quality are currently insufficiently sensitive for some pathogens, notably viruses and parasites (e.g., Cryptosporidium). Therefore, in the absence of evidence from an analytical study to demonstrate an association, the link between illness and water can only be classified as probable or possible. However, total coliform counts are good indicators of faecal contamination with other pathogens (e.g., rotaviruses, E. coli, Campylobacter) [43]. In conclusion, the waterborne outbreak described in the present study was most probably caused by faecal contamination of source A by groundwater, resulting from deficiencies in the maintenance of the water distribution system. Corrective measures were implemented based on the results of the investigation. To our knowledge, this is the first documented waterborne outbreak involving human C. coli infection. In-depth environmental investigations to identify the origin of water contamination are important in order to better understand microbial transmission from the environment and to implement appropriate controls and preventive measures. In addition, surveys on vehicles and vectors that transmit Campylobacter between hosts are crucial in order to better understand the epidemiology of these organisms.

How many people were ill?

295

A large multi-pathogen waterborne community outbreak linked to faecal contamination of a groundwater system, France, 2000

A large waterborne outbreak of infection that occurred during August 2000 in a local community in France was investigated initially via a rapid survey of visits to local physicians. A retrospective cohort study was then conducted on a random cluster sample of residents. Of 709 residents interviewed, 202 (28.5%) were definite cases (at least three liquid stools ℠day or vomiting) and 62 (8.7%) were probable cases (less than three liquid stools ℠day or abdominal pain). Those who had drunk tap water had a threefold increased risk for illness (95% CI 2.4–4.0). The risk increased with the amount of water consumed (chi-square trend: p < 0.0001). Bacteriological analyses of stools were performed for 35 patients and virological analyses for 24 patients. Campylobacter coli, group A rotavirus and norovirus were detected in 31.5%, 71.0% and 21% of samples, respectively. An extensive environmental investigation concluded that a groundwater source to this community had probably been contaminated by agricultural run-off, and a failure in the chlorination system was identified. This is the first documented waterborne outbreak of infection involving human C. coli infections. A better understanding of the factors influencing campylobacter transmission between hosts is required. Keywords Campylobacter coli, epidemiology, gastroenteritis, norovirus, rotavirus, waterborne outbreak INTRODUCTION Contaminated drinking water causes extensive outbreaks of illness because of the large number of people served by water supply facilities. Detection of such outbreaks requires the identification of an increase in illness (usually the rate of gastrointestinal disease) in the exposed population and confirmation that water was the route of transmission [1]. Many countries have routine reporting systems for detecting foodborne and waterborne outbreaks of infection. However, in routine practice it is not easy to detect an increase in clinical cases and to link this increase to waterborne transmission; thus waterborne outbreaks are often unrecognised and underestimated [2]. Microbiological examination of water is more complex than examination of stools, and there is often a failure to detect pathogens in water. Norovirus, Cryptosporidium parvum, Giardia intestinalis and Campylobacter jejuni are the pathogens identified most frequently [3]. The present study describes a large waterborne community outbreak of infection with multiple pathogens, including Campylobacter coli. METHODS AND MATERIALS Background On 23 August 2000, a general practitioner (GP) informed the local district health department that 16 cases of acute gastroenteritis (AGE) had occurred among residents of a holiday camp in the Gourdon community since 21 August. Gourdon, which is located in the south-west of France, has a population of 4888 inhabitants, which increases by 50% in the July– August tourist season. Investigations at the holiday camp concluded that foodborne transmission was very unlikely, but found that tap water was highly contaminated with faecal coliforms. Preliminary information collected from physicians in Gourdon indicated an increase of the number of consultations for AGE. Gourdon tap water is supplied by two underground water facilities: facility A serving 3800 water consumers in Gourdon; and facility B serving 1200 water consumers in Gourdon and the surrounding communities. Water treatment in facility A consists of chlorination, which involves two automatic pumps working in rotation. The water then passes through two semi in-ground reservoir tanks (1100 m3 and 200 m3, respectively) before distribution. Facility B water is mixed with a fraction of facility A water in another 500 m3 semi in-ground reservoir tank before being distributed. This mixing with the treated water from facility A is intended to disinfect the water from facility B. Since the increase in AGE was noted only in Gourdon, contaminated tap water from facility A was suspected to be the origin of the outbreak. Therefore, the population of Gourdon was informed on 25 August to avoid the consumption of tap water or to boil it for 5 min. Survey of gastroenteritis-related medical consultations To confirm the outbreak, the 11 GPs in Gourdon and the local hospital emergency department were requested to provide the number of total daily visits and the number of cases of gastroenteritis cases occurring between 1 August and 11 September 2000. Retrospective cohort study A retrospective cohort study was performed on a sample of Gourdon households between 5 and 15 September 2000. Domestic households were selected randomly from the telephone directory. A definite case was defined as a Gourdon resident who had at least three liquid stools ℠day or vomiting between 1 August and 3 September, and who had been present in Gourdon for ‡ 1 day between 1 and 31 August; if less than three liquid stools ℠day or abdominal pain had occurred, the case was classified as probable. Exposure was defined as the consumption of tap water before 25 August 2000. A standardised telephone questionnaire was completed to obtain demographic data (gender, date of birth), food consumption (commercially prepared dishes, sea-food, raw milk, pastry), tap water consumption (before and after 25 August, daily number of glasses of water, consumption of boiled tap water), symptoms of gastroenteritis (date of onset, number of liquid stools ℠day, vomiting, nausea, abdominal pain, fever), medical care (in- or out-patient) and consequences on daily activities (sick leave, bed confinement) for each member of the selected household. Under the hypothesis that 60% of the population consumed tap water, an estimated 550 people were required for inclusion in the study to identify a relative risk (RR) of ‡ 3, with a 5% a risk and a power of 80%. Based on an average of two individuals per household (INSERM 1999 population census), 300 households were required for inclusion in the survey. The association between AGE and the consumption of tap water was assessed by calculating the risk ratio (RR) and 95% CI, taking into account a household cluster design effect. A dose-effect relationship was evaluated by analysing the RR trend for different categories of daily number of glasses of tap water consumed (chi-square trend). Proportions were compared using the corrected chi-square test, and means were compared using the Student test or Mann and Witney test. Data were analysed with Epi-Info 6.04c (CDC, Atlanta, GA, USA). The gender and age distribution of the cohort population were compared with that of the general population of Gourdon (INSERM 1999 population census) to assess whether it was representative. Under the hypothesis of representation, the total number of cases of gastroenteritis was estimated by applying the attack rate to the total population of Gourdon, including the tourist population (data provided by the Gourdon tourism office). The most likely period for contamination of the water distribution system was based on the epidemic curve and the minimal incubation period of the pathogens involved. Microbiological analyses of stools Stool samples were examined for the presence of Salmonella spp., Shigella spp., Staphylococcus aureus, Campylobacter spp., Yersinia enterocolitica, Escherichia coli, enterococci, rotavirus groups A and C, astrovirus, calicivirus, adenovirus types 40 and 41, enterovirus, hepatitis A virus and Cryptosporidum. Viruses were detected using an enzyme immunoassay for group A rotavirus, astrovirus and adenovirus 40 ℠41, and by RT-PCR for calicivirus, group C rotavirus, enterovirus and hepatitis A virus [4–6]. Campylobacter isolates were sent to the National Reference Centre for Campylobacter for identification and fingerprinting. Species identification was by means of phenotypic tests (API Campy; bioMe´rieux, Lyon, France), with use of PCR to differentiate Campylobacter jejuni from Camplylobacter coli [7]. Random amplified polymorphic DNA fingerprinting was performed with primer 3881, 5¢-AACGCGCAAC. Environmental investigation An environmental and microbiological investigation of the water distribution system was performed. The inspection concerned the water catchments, the water distribution and sewage system, agricultural practices and private households around the sources. Bacteriological (Escherichia coli, enterococci, faecal streptococci, thermotolerant coliforms, sulphite-reducing clostridia) and chemical (chlorine) analyses were performed on tap water and on the water sampled from the two groundwater sources. After 23 August, the water chlorine concentration and the presence of coliform bacteria were monitored daily at several sites of the water system before and after the resort treatment. Examinations for viruses (group A rotavirus, astrovirus, norovirus, enterovirus, hepatitis A virus) and Campylobacter spp. were undertaken on two 10-mL tap water samples collected inhouse on 23 August, and on two 10-mL samples from groundwater sources A and B on 25 August. Viruses were detected by RT-PCR, followed by hybridisation with a specific primer to improve sensitivity, while C. jejuni and C. coli were detected by PCR with specific primers. Group A rotavirus isolates detected in stools and in water were sequenced using Beg 9 and End 9 primers [8] and compared using the Infobiogene GCG software (http://www.infobiogen.fr). Bacteriological (revived 36C, revived 22C, thermotolerant coliforms, streptococci and faecal enterococci), physical (temperature, conductivity) and chemical (oxygen consumption, ammonium, nitrates, phosphorus) parameters from source A and the river were analysed. Trends were compared for an 8-week period between 21 December 2000 and 7 February 2001. RESULTS Epidemiology Between 1 August and 11 September 2000, 7104 Gourdon residents consulted a GP or the local hospital emergency department. Between 24 and 28 August, gastroenteritis accounted for 44% (479 ℠1097) of all medical visits, compared with 6% (126 ℠2185) during the first 15 days of August (Fig. 1). Among the 498 households contacted for the household survey, 198 were excluded (103 did not respond, 49 did not fulfil the inclusion criteria, 43 refused to participate, and three for unknown reasons). The questionnaire was completed for 300 households (709 individuals). Data were missing concerning gender for one individual and age for 11 individuals. Overall, 331 (46.7%; 95% CI 44.5–49.0) of respondents were male and 377 (53.2%; 95% CI 50.9–55.5) were female. The median age was 44 years (range < 1–94 years). The age and gender distribution of the sample was similar to that of the general population. During the outbreak period, 264 of the 709 respondents (attack rate (AR) 37.2%) had been ill, with 202 definite cases (AR 28.5%; 95% CI 24.6– 32.4) and 62 probable cases (AR 8.7%; 95% CI 6.5– 11.0). The AR was higher in females (31.8%) than in males (24.8%), and was highest in children aged < 6 years (42.1%, Table 1). The epidemic curve (Fig. 2) suggests that the outbreak started between 14 and 19 August. The shape of the epidemic curve in the retrospective cohort study was very similar to that based on the number of visits to GPs for gastroenteritis. Patients complained mainly of diarrhoea, abdominal pain and nausea (Table 2). No cases of bloody diarrhoea were notified. The mean duration of symptoms was 4.5 days (median 3, range 1–30 days). Among definite cases, 52% had visited a GP, 42% had taken sick leave for ‡ 1 day, 32% had required bed rest, and 2.9% (six patients) had been hospitalised. In total, 336 (47.4%) respondents had drunk tap water before 25 August, whereas 646 (91.1%) did not drink tap water after 25 August. The risk for illness, taking into account the sample cluster design effect of 1.4, was three-fold greater (95% CI 2.2–4.1) for those who had drunk tap water before 25 August, and increased with the amount of water consumed (chi-square trend: p < 0.0001; Table 3). The risk of gastroenteritis associated with drinking tap water did not change substantially after adjustment for age (RRMH 3.3, 95% CI 2.5–4.4), and there was no modification effect according to age. Food consumption was not associated with illness. Applying the AR of 37% to the total population of Gourdon exposed to the water supply (4888 residents + 2200 tourists), the number of individuals affected by this outbreak was estimated to be 2600 (95% CI 2400–2900), assuming an exposure to tap water and a risk for gastroenteritis among the tourists similar to that among the inhabitants. Microbiological stool analyses Among the 35 stool samples analysed for bacteria, 11 (31.5%) were positive for C. coli. Twenty-four of the 35 samples were also tested for viruses: rotavirus A was detected in 17 (71.0%) and norovirus in five (21.0%) samples, while nine (37.5%) samples were co-infected (four C. coli + rotavirus; one C. coli + norovirus; three rotavirus + norovirus; and one C. coli + rotavirus + norovirus). Of the 11 C. coli isolates, six were characterised as biotype 2, and were identical following RAPD analysis, while the remaining five were of biotype 1. Rotavirus typing by RTPCR identified a single genotype P-(8), G1. Molecular characterisation of the norovirus isolates identified two different patterns, with three isolates being identical and close to the Saragota strain (genogroup I) and two close to the new GGIIb variant (genogroup II). Environmental investigation All of the water samples from source B were negative, regardless of the site of the distribution system tested. Analyses of the water originating from source A, sampled after leaving the water-works on 23 and 25 August, revealed high concentrations of E. coli (2263 / 100 mL), enterococci (415 / 100 mL), total coliforms (90 / 100 mL), faecal streptococci (136 / 100 mL) and sulphite-reducing clostridia (10 / 100 mL), and an absence of chlorination. Rotavirus A genotype G1 was detected in the crude water sampled from source A. The analyses for enterovirus, astrovirus, norovirus, hepatitis A virus and Campylobacter were negative. Source A was situated in a valley containing a river, below a hamlet of 15 dwellings, including farms breeding cattle, pork and sheep. The source was not supposed to be supplied by the river. The river resulted from the convergence of about ten different streams, with one originating close to a slaughterhouse, and was exposed to agricultural run-off and manure effluent; sheep and cattle grazed in meadows bordering the river, and a large amount of sewage from water treatment facilities was sprayed on to land close to the river. Cowsheds were close to the water catchment area. The immediate protected area around source A was too small and littered with building rubbish. Air vents without protection in the door made the source accessible to animals and insects. Bacteriological analyses of the water samples from source A and the river revealed that the contamination was of both human and animal origin, with a higher concentration in the river. The levels and trends of the chemical parameters of source A were similar to those of the river, strongly suggesting that the river was connected to source A (Fig. 3). A retrospective review of routine data collected between 1991 and 2000 showed irregularities in the maintenance of the water distribution system and the water treatment plant, particularly during the month of August. In the hamlet, the sewage system of the 15 households did not comply with sanitary rules. All but three ejected waste either into a cesspool or superficially without treatment. Three regions of gene 9 of the rotavirus detected in stools and water were compared. Sequencing showed that the strains were closely related but not identical. Assuming a likely onset of the outbreak between 14 and 19 August, and taking into account a minimum incubation period of 2 days for rotavirus A, norovirus and C. coli, the contamination of the water distribution system probably occurred between 12 and 17 August. Contamination probably persisted until 28 August, as intensive chlorination, beginning on 25 August, was effective only from 28 August onwards (Fig. 2). DISCUSSION The present descriptive and retrospective cohort studies confirmed that the outbreak of infection in the community of Gourdon was related to the consumption of tap water and affected 2600 individuals. The outbreak involved multiple pathogens, with group A rotavirus and C. coli being the most frequent, followed by norovirus. Results of environmental investigations suggested that water contamination was related to the contamination of the source, combined with a failure of the water chlorination system. The association with tap water consumption, the dose-effect relationship, and the fact that multiple pathogens were involved, were all strongly in favour of tap water as the origin of infection. Moreover, the number of cases decreased dramatically after the water restriction advice and the implementation of control measures on the water supply. No food products explored in the study were associated with illness. An AR of 14.5% and 31% was observed among adults and children aged < 6 years, respectively, who reported that they had not consumed tap water. This relatively high rate could be explained by several phenomena. Many individuals may have consumed tap water without being aware of it, e.g., in drinks or ice cubes consumed outside of the home, via raw fruits and vegetables, dishes prepared with unboiled water, or while cleaning their teeth. It can be hypothesised that people who became ill without being aware of having consumed tap water were exposed to small amounts of water and had been subjected to a small infective dose. In addition, children, more than adults, may have been contaminated through person-to-person transmission (e.g., family contact), which is the most common route of transmission for rotavirus and calicivirus. Waterborne contamination can affect large numbers of individuals [3,9–11]. In the present study, assuming a 50% increase in the population during the tourist season, the number of affected people was estimated to be between 2400 and 2900. The cohort study showed that almost half (46.5%) of the cases consulted a GP. Applying this proportion to the estimated number of cases yielded an estimated 1116–1349 ill individuals who consulted a GP for gastroenteritis, similar to the number of consultations for gastroenteritis actually reported by the GPs (n = 1037). The stool sample analyses revealed the presence of multiple pathogens (rotavirus A, norovirus genogroups I and II and C. coli) as single or co-infections. Except for C. coli, these pathogens have been implicated in numerous waterborne outbreaks, suggesting an influx of sewage into the water supply system [10,12–18]. Rotaviruses are known to be responsible for large epidemics during winter seasons, affecting particularly children aged < 3 years. These epidemics are caused by the faecal route of contamination (personto-person transmission). The high proportion of stools positive for rotavirus (71%) is uncommon in waterborne outbreaks and probably reflects massive contamination [15,16,19]. Different molecular sequences of rotavirus A genotype G1 found in stools and water indicated a human faecal source of infection. In the present study, contamination of water by sewage could be linked with the large amount of sewage from the water purification plant sprayed on to land close to the river and the water catchments. In this outbreak, C. coli, which was identified in 32% of stools, was the second most frequent pathogen. To our knowledge, human C. coli contamination has not been described previously as a causative agent in waterborne outbreaks, unlike C. jejuni [11,20–25], although C. coli has been isolated from water [26]. The main symptoms observed (diarrhoea, abdominal pain, nausea) and the mean duration of the disease (4.5 days) were compatible with a Campylobacter infection. No bloody diarrhoea was reported, although bloody diarrhoea is not constant in Campylobacter infections [27]. The high proportion of individuals who had to take sick leave, and the fact that six individuals were hospitalised, suggested a relative severity of symptoms, which is much more compatible with Campylobacter infection than with virus infections. Among the hospitalised patients, 60% were infected with C. coli. Among patients whose stools were analysed and whose age was known, cases infected only with rotavirus were younger (mean age 49 years; range, 5 months to 95 years) than cases infected only with Campylobacter (mean age 66 years; range 23–88 years) or co-infected with rotavirus plus Campylobacter (mean age 69 years; range 2–91 years). Moreover, the proportion (52%) of medical visits was unusually high compared with the proportion of visits observed during other waterborne outbreaks [13,16,28]. Campylobacter needs specific atmospheric conditions for growth and survival and its detection can be very difficult [29,30]. Because of a rapid physiological transformation to a viable, though non-cultivable, state, it has been suggested that conventional culture methods often fail [31]. Nevertheless, in the present study, the failure to detect C. coli in water, and the detection of noroviruses in stools, could be caused by the small (10 mL) water samples examined. Ha¨nninen et al. [30] proposed that the volume used for detection of suspected pathogens in drinking water should be at least 8–10 L. Groundwater is considered widely to be microbiologically clean, and is usually used for drinking without treatment. However, groundwater can be contaminated by the environment [26,32]. Stanley et al. [33] showed that Campylobacter can occur in groundwater, with hydrological evidence suggesting that the source of contamination in their study was a dairy farm situated within the hydrological catchment area of the groundwater, which was confirmed when identical strains of C. jejuni were isolated from groundwater and the dairy herd. In the present study, Campylobacter was not identified from groundwater, and dairy herds and other domestic animals in farms around the source were not investigated. However, results of bacterial, chemical and physical water analyses suggested strongly that a river contaminated by agricultural run-off from surrounding sheep and cattle grazing meadows was contaminating the source. In addition, the immediate area of the source was insufficiently protected and vulnerable to the environment. Wild birds, rodents, pigs and cows are all hosts for C. coli and could contaminate surface water easily [26,27,34,35]. Campylobacters are sensitive to many environmental factors and to chlorine [33,36,37], but a failure of the chlorination system could have resulted in the presence of a high concentration of C. coli. Waterborne outbreaks are often caused by several factors that act together [18,20,23,38]. In some outbreaks, heavy rainfall has been responsible for increasing agricultural run-off into rivers, although this did not occur before the outbreak in Gourdon. This underlies the need for increased frequency of routine water quality monitoring, chlorination of the water supply according to environmental conditions, and implementation of preventive measures. This outbreak raised the question of the delay in raising the alert. An outbreak can be recognised by: (i) the poor quality of water detected during regular controls; (ii) complaints about water quality; (iii) an incident in the supply and ℠or sewage water system during maintenance work; and (iv) an increased number of cases of gastroenteritis. In most instances, waterborne outbreaks are suspected when there are already a large number of declared cases [3,12,18,28,38,39]. In the present outbreak, the initial warning was given by a GP who notified officials 5–10 days after the assumed date of onset. This highlights the role of GPs in the detection of an unusual event, particularly during the summer, when an increase in acute gastroenteritis is less expected than during winter [40,41]. In addition, a routine water quality monitoring system able to detect contamination of drinking water before the occurrence of illness and complaints about the water quality should be used as an early warning source. The retrospective cohort study of Gourdon residents allowed the scale and severity of the outbreak to be quantified precisely and the hypothesis that tap water was the vehicle of the outbreak to be tested. When provided with a cohort sample representative of the study population, the number of people affected can be estimated. However, as the exposure of the population was high, a relatively large sample size was needed to test the association between the disease and the exposure. Because of misclassification of exposure (unrecognised consumption of tap water), the attributable risk is probably underestimated. In the absence of obvious alternative explanations, water contamination is the most likely common cause of an explosive outbreak of AGE occurring in a large population. Tillett et al. [42] proposed categorising levels of evidence for waterborne outbreaks. A descriptive study suggesting a water-related outbreak, together with the identification of the same pathogen in stools and water, should be sufficient to conclude that an outbreak is associated strongly with water consumption. Faecal indicators of water quality are currently insufficiently sensitive for some pathogens, notably viruses and parasites (e.g., Cryptosporidium). Therefore, in the absence of evidence from an analytical study to demonstrate an association, the link between illness and water can only be classified as probable or possible. However, total coliform counts are good indicators of faecal contamination with other pathogens (e.g., rotaviruses, E. coli, Campylobacter) [43]. In conclusion, the waterborne outbreak described in the present study was most probably caused by faecal contamination of source A by groundwater, resulting from deficiencies in the maintenance of the water distribution system. Corrective measures were implemented based on the results of the investigation. To our knowledge, this is the first documented waterborne outbreak involving human C. coli infection. In-depth environmental investigations to identify the origin of water contamination are important in order to better understand microbial transmission from the environment and to implement appropriate controls and preventive measures. In addition, surveys on vehicles and vectors that transmit Campylobacter between hosts are crucial in order to better understand the epidemiology of these organisms.

How many people were hospitalized?

296

A large multi-pathogen waterborne community outbreak linked to faecal contamination of a groundwater system, France, 2000

A large waterborne outbreak of infection that occurred during August 2000 in a local community in France was investigated initially via a rapid survey of visits to local physicians. A retrospective cohort study was then conducted on a random cluster sample of residents. Of 709 residents interviewed, 202 (28.5%) were definite cases (at least three liquid stools ℠day or vomiting) and 62 (8.7%) were probable cases (less than three liquid stools ℠day or abdominal pain). Those who had drunk tap water had a threefold increased risk for illness (95% CI 2.4–4.0). The risk increased with the amount of water consumed (chi-square trend: p < 0.0001). Bacteriological analyses of stools were performed for 35 patients and virological analyses for 24 patients. Campylobacter coli, group A rotavirus and norovirus were detected in 31.5%, 71.0% and 21% of samples, respectively. An extensive environmental investigation concluded that a groundwater source to this community had probably been contaminated by agricultural run-off, and a failure in the chlorination system was identified. This is the first documented waterborne outbreak of infection involving human C. coli infections. A better understanding of the factors influencing campylobacter transmission between hosts is required. Keywords Campylobacter coli, epidemiology, gastroenteritis, norovirus, rotavirus, waterborne outbreak INTRODUCTION Contaminated drinking water causes extensive outbreaks of illness because of the large number of people served by water supply facilities. Detection of such outbreaks requires the identification of an increase in illness (usually the rate of gastrointestinal disease) in the exposed population and confirmation that water was the route of transmission [1]. Many countries have routine reporting systems for detecting foodborne and waterborne outbreaks of infection. However, in routine practice it is not easy to detect an increase in clinical cases and to link this increase to waterborne transmission; thus waterborne outbreaks are often unrecognised and underestimated [2]. Microbiological examination of water is more complex than examination of stools, and there is often a failure to detect pathogens in water. Norovirus, Cryptosporidium parvum, Giardia intestinalis and Campylobacter jejuni are the pathogens identified most frequently [3]. The present study describes a large waterborne community outbreak of infection with multiple pathogens, including Campylobacter coli. METHODS AND MATERIALS Background On 23 August 2000, a general practitioner (GP) informed the local district health department that 16 cases of acute gastroenteritis (AGE) had occurred among residents of a holiday camp in the Gourdon community since 21 August. Gourdon, which is located in the south-west of France, has a population of 4888 inhabitants, which increases by 50% in the July– August tourist season. Investigations at the holiday camp concluded that foodborne transmission was very unlikely, but found that tap water was highly contaminated with faecal coliforms. Preliminary information collected from physicians in Gourdon indicated an increase of the number of consultations for AGE. Gourdon tap water is supplied by two underground water facilities: facility A serving 3800 water consumers in Gourdon; and facility B serving 1200 water consumers in Gourdon and the surrounding communities. Water treatment in facility A consists of chlorination, which involves two automatic pumps working in rotation. The water then passes through two semi in-ground reservoir tanks (1100 m3 and 200 m3, respectively) before distribution. Facility B water is mixed with a fraction of facility A water in another 500 m3 semi in-ground reservoir tank before being distributed. This mixing with the treated water from facility A is intended to disinfect the water from facility B. Since the increase in AGE was noted only in Gourdon, contaminated tap water from facility A was suspected to be the origin of the outbreak. Therefore, the population of Gourdon was informed on 25 August to avoid the consumption of tap water or to boil it for 5 min. Survey of gastroenteritis-related medical consultations To confirm the outbreak, the 11 GPs in Gourdon and the local hospital emergency department were requested to provide the number of total daily visits and the number of cases of gastroenteritis cases occurring between 1 August and 11 September 2000. Retrospective cohort study A retrospective cohort study was performed on a sample of Gourdon households between 5 and 15 September 2000. Domestic households were selected randomly from the telephone directory. A definite case was defined as a Gourdon resident who had at least three liquid stools ℠day or vomiting between 1 August and 3 September, and who had been present in Gourdon for ‡ 1 day between 1 and 31 August; if less than three liquid stools ℠day or abdominal pain had occurred, the case was classified as probable. Exposure was defined as the consumption of tap water before 25 August 2000. A standardised telephone questionnaire was completed to obtain demographic data (gender, date of birth), food consumption (commercially prepared dishes, sea-food, raw milk, pastry), tap water consumption (before and after 25 August, daily number of glasses of water, consumption of boiled tap water), symptoms of gastroenteritis (date of onset, number of liquid stools ℠day, vomiting, nausea, abdominal pain, fever), medical care (in- or out-patient) and consequences on daily activities (sick leave, bed confinement) for each member of the selected household. Under the hypothesis that 60% of the population consumed tap water, an estimated 550 people were required for inclusion in the study to identify a relative risk (RR) of ‡ 3, with a 5% a risk and a power of 80%. Based on an average of two individuals per household (INSERM 1999 population census), 300 households were required for inclusion in the survey. The association between AGE and the consumption of tap water was assessed by calculating the risk ratio (RR) and 95% CI, taking into account a household cluster design effect. A dose-effect relationship was evaluated by analysing the RR trend for different categories of daily number of glasses of tap water consumed (chi-square trend). Proportions were compared using the corrected chi-square test, and means were compared using the Student test or Mann and Witney test. Data were analysed with Epi-Info 6.04c (CDC, Atlanta, GA, USA). The gender and age distribution of the cohort population were compared with that of the general population of Gourdon (INSERM 1999 population census) to assess whether it was representative. Under the hypothesis of representation, the total number of cases of gastroenteritis was estimated by applying the attack rate to the total population of Gourdon, including the tourist population (data provided by the Gourdon tourism office). The most likely period for contamination of the water distribution system was based on the epidemic curve and the minimal incubation period of the pathogens involved. Microbiological analyses of stools Stool samples were examined for the presence of Salmonella spp., Shigella spp., Staphylococcus aureus, Campylobacter spp., Yersinia enterocolitica, Escherichia coli, enterococci, rotavirus groups A and C, astrovirus, calicivirus, adenovirus types 40 and 41, enterovirus, hepatitis A virus and Cryptosporidum. Viruses were detected using an enzyme immunoassay for group A rotavirus, astrovirus and adenovirus 40 ℠41, and by RT-PCR for calicivirus, group C rotavirus, enterovirus and hepatitis A virus [4–6]. Campylobacter isolates were sent to the National Reference Centre for Campylobacter for identification and fingerprinting. Species identification was by means of phenotypic tests (API Campy; bioMe´rieux, Lyon, France), with use of PCR to differentiate Campylobacter jejuni from Camplylobacter coli [7]. Random amplified polymorphic DNA fingerprinting was performed with primer 3881, 5¢-AACGCGCAAC. Environmental investigation An environmental and microbiological investigation of the water distribution system was performed. The inspection concerned the water catchments, the water distribution and sewage system, agricultural practices and private households around the sources. Bacteriological (Escherichia coli, enterococci, faecal streptococci, thermotolerant coliforms, sulphite-reducing clostridia) and chemical (chlorine) analyses were performed on tap water and on the water sampled from the two groundwater sources. After 23 August, the water chlorine concentration and the presence of coliform bacteria were monitored daily at several sites of the water system before and after the resort treatment. Examinations for viruses (group A rotavirus, astrovirus, norovirus, enterovirus, hepatitis A virus) and Campylobacter spp. were undertaken on two 10-mL tap water samples collected inhouse on 23 August, and on two 10-mL samples from groundwater sources A and B on 25 August. Viruses were detected by RT-PCR, followed by hybridisation with a specific primer to improve sensitivity, while C. jejuni and C. coli were detected by PCR with specific primers. Group A rotavirus isolates detected in stools and in water were sequenced using Beg 9 and End 9 primers [8] and compared using the Infobiogene GCG software (http://www.infobiogen.fr). Bacteriological (revived 36C, revived 22C, thermotolerant coliforms, streptococci and faecal enterococci), physical (temperature, conductivity) and chemical (oxygen consumption, ammonium, nitrates, phosphorus) parameters from source A and the river were analysed. Trends were compared for an 8-week period between 21 December 2000 and 7 February 2001. RESULTS Epidemiology Between 1 August and 11 September 2000, 7104 Gourdon residents consulted a GP or the local hospital emergency department. Between 24 and 28 August, gastroenteritis accounted for 44% (479 ℠1097) of all medical visits, compared with 6% (126 ℠2185) during the first 15 days of August (Fig. 1). Among the 498 households contacted for the household survey, 198 were excluded (103 did not respond, 49 did not fulfil the inclusion criteria, 43 refused to participate, and three for unknown reasons). The questionnaire was completed for 300 households (709 individuals). Data were missing concerning gender for one individual and age for 11 individuals. Overall, 331 (46.7%; 95% CI 44.5–49.0) of respondents were male and 377 (53.2%; 95% CI 50.9–55.5) were female. The median age was 44 years (range < 1–94 years). The age and gender distribution of the sample was similar to that of the general population. During the outbreak period, 264 of the 709 respondents (attack rate (AR) 37.2%) had been ill, with 202 definite cases (AR 28.5%; 95% CI 24.6– 32.4) and 62 probable cases (AR 8.7%; 95% CI 6.5– 11.0). The AR was higher in females (31.8%) than in males (24.8%), and was highest in children aged < 6 years (42.1%, Table 1). The epidemic curve (Fig. 2) suggests that the outbreak started between 14 and 19 August. The shape of the epidemic curve in the retrospective cohort study was very similar to that based on the number of visits to GPs for gastroenteritis. Patients complained mainly of diarrhoea, abdominal pain and nausea (Table 2). No cases of bloody diarrhoea were notified. The mean duration of symptoms was 4.5 days (median 3, range 1–30 days). Among definite cases, 52% had visited a GP, 42% had taken sick leave for ‡ 1 day, 32% had required bed rest, and 2.9% (six patients) had been hospitalised. In total, 336 (47.4%) respondents had drunk tap water before 25 August, whereas 646 (91.1%) did not drink tap water after 25 August. The risk for illness, taking into account the sample cluster design effect of 1.4, was three-fold greater (95% CI 2.2–4.1) for those who had drunk tap water before 25 August, and increased with the amount of water consumed (chi-square trend: p < 0.0001; Table 3). The risk of gastroenteritis associated with drinking tap water did not change substantially after adjustment for age (RRMH 3.3, 95% CI 2.5–4.4), and there was no modification effect according to age. Food consumption was not associated with illness. Applying the AR of 37% to the total population of Gourdon exposed to the water supply (4888 residents + 2200 tourists), the number of individuals affected by this outbreak was estimated to be 2600 (95% CI 2400–2900), assuming an exposure to tap water and a risk for gastroenteritis among the tourists similar to that among the inhabitants. Microbiological stool analyses Among the 35 stool samples analysed for bacteria, 11 (31.5%) were positive for C. coli. Twenty-four of the 35 samples were also tested for viruses: rotavirus A was detected in 17 (71.0%) and norovirus in five (21.0%) samples, while nine (37.5%) samples were co-infected (four C. coli + rotavirus; one C. coli + norovirus; three rotavirus + norovirus; and one C. coli + rotavirus + norovirus). Of the 11 C. coli isolates, six were characterised as biotype 2, and were identical following RAPD analysis, while the remaining five were of biotype 1. Rotavirus typing by RTPCR identified a single genotype P-(8), G1. Molecular characterisation of the norovirus isolates identified two different patterns, with three isolates being identical and close to the Saragota strain (genogroup I) and two close to the new GGIIb variant (genogroup II). Environmental investigation All of the water samples from source B were negative, regardless of the site of the distribution system tested. Analyses of the water originating from source A, sampled after leaving the water-works on 23 and 25 August, revealed high concentrations of E. coli (2263 / 100 mL), enterococci (415 / 100 mL), total coliforms (90 / 100 mL), faecal streptococci (136 / 100 mL) and sulphite-reducing clostridia (10 / 100 mL), and an absence of chlorination. Rotavirus A genotype G1 was detected in the crude water sampled from source A. The analyses for enterovirus, astrovirus, norovirus, hepatitis A virus and Campylobacter were negative. Source A was situated in a valley containing a river, below a hamlet of 15 dwellings, including farms breeding cattle, pork and sheep. The source was not supposed to be supplied by the river. The river resulted from the convergence of about ten different streams, with one originating close to a slaughterhouse, and was exposed to agricultural run-off and manure effluent; sheep and cattle grazed in meadows bordering the river, and a large amount of sewage from water treatment facilities was sprayed on to land close to the river. Cowsheds were close to the water catchment area. The immediate protected area around source A was too small and littered with building rubbish. Air vents without protection in the door made the source accessible to animals and insects. Bacteriological analyses of the water samples from source A and the river revealed that the contamination was of both human and animal origin, with a higher concentration in the river. The levels and trends of the chemical parameters of source A were similar to those of the river, strongly suggesting that the river was connected to source A (Fig. 3). A retrospective review of routine data collected between 1991 and 2000 showed irregularities in the maintenance of the water distribution system and the water treatment plant, particularly during the month of August. In the hamlet, the sewage system of the 15 households did not comply with sanitary rules. All but three ejected waste either into a cesspool or superficially without treatment. Three regions of gene 9 of the rotavirus detected in stools and water were compared. Sequencing showed that the strains were closely related but not identical. Assuming a likely onset of the outbreak between 14 and 19 August, and taking into account a minimum incubation period of 2 days for rotavirus A, norovirus and C. coli, the contamination of the water distribution system probably occurred between 12 and 17 August. Contamination probably persisted until 28 August, as intensive chlorination, beginning on 25 August, was effective only from 28 August onwards (Fig. 2). DISCUSSION The present descriptive and retrospective cohort studies confirmed that the outbreak of infection in the community of Gourdon was related to the consumption of tap water and affected 2600 individuals. The outbreak involved multiple pathogens, with group A rotavirus and C. coli being the most frequent, followed by norovirus. Results of environmental investigations suggested that water contamination was related to the contamination of the source, combined with a failure of the water chlorination system. The association with tap water consumption, the dose-effect relationship, and the fact that multiple pathogens were involved, were all strongly in favour of tap water as the origin of infection. Moreover, the number of cases decreased dramatically after the water restriction advice and the implementation of control measures on the water supply. No food products explored in the study were associated with illness. An AR of 14.5% and 31% was observed among adults and children aged < 6 years, respectively, who reported that they had not consumed tap water. This relatively high rate could be explained by several phenomena. Many individuals may have consumed tap water without being aware of it, e.g., in drinks or ice cubes consumed outside of the home, via raw fruits and vegetables, dishes prepared with unboiled water, or while cleaning their teeth. It can be hypothesised that people who became ill without being aware of having consumed tap water were exposed to small amounts of water and had been subjected to a small infective dose. In addition, children, more than adults, may have been contaminated through person-to-person transmission (e.g., family contact), which is the most common route of transmission for rotavirus and calicivirus. Waterborne contamination can affect large numbers of individuals [3,9–11]. In the present study, assuming a 50% increase in the population during the tourist season, the number of affected people was estimated to be between 2400 and 2900. The cohort study showed that almost half (46.5%) of the cases consulted a GP. Applying this proportion to the estimated number of cases yielded an estimated 1116–1349 ill individuals who consulted a GP for gastroenteritis, similar to the number of consultations for gastroenteritis actually reported by the GPs (n = 1037). The stool sample analyses revealed the presence of multiple pathogens (rotavirus A, norovirus genogroups I and II and C. coli) as single or co-infections. Except for C. coli, these pathogens have been implicated in numerous waterborne outbreaks, suggesting an influx of sewage into the water supply system [10,12–18]. Rotaviruses are known to be responsible for large epidemics during winter seasons, affecting particularly children aged < 3 years. These epidemics are caused by the faecal route of contamination (personto-person transmission). The high proportion of stools positive for rotavirus (71%) is uncommon in waterborne outbreaks and probably reflects massive contamination [15,16,19]. Different molecular sequences of rotavirus A genotype G1 found in stools and water indicated a human faecal source of infection. In the present study, contamination of water by sewage could be linked with the large amount of sewage from the water purification plant sprayed on to land close to the river and the water catchments. In this outbreak, C. coli, which was identified in 32% of stools, was the second most frequent pathogen. To our knowledge, human C. coli contamination has not been described previously as a causative agent in waterborne outbreaks, unlike C. jejuni [11,20–25], although C. coli has been isolated from water [26]. The main symptoms observed (diarrhoea, abdominal pain, nausea) and the mean duration of the disease (4.5 days) were compatible with a Campylobacter infection. No bloody diarrhoea was reported, although bloody diarrhoea is not constant in Campylobacter infections [27]. The high proportion of individuals who had to take sick leave, and the fact that six individuals were hospitalised, suggested a relative severity of symptoms, which is much more compatible with Campylobacter infection than with virus infections. Among the hospitalised patients, 60% were infected with C. coli. Among patients whose stools were analysed and whose age was known, cases infected only with rotavirus were younger (mean age 49 years; range, 5 months to 95 years) than cases infected only with Campylobacter (mean age 66 years; range 23–88 years) or co-infected with rotavirus plus Campylobacter (mean age 69 years; range 2–91 years). Moreover, the proportion (52%) of medical visits was unusually high compared with the proportion of visits observed during other waterborne outbreaks [13,16,28]. Campylobacter needs specific atmospheric conditions for growth and survival and its detection can be very difficult [29,30]. Because of a rapid physiological transformation to a viable, though non-cultivable, state, it has been suggested that conventional culture methods often fail [31]. Nevertheless, in the present study, the failure to detect C. coli in water, and the detection of noroviruses in stools, could be caused by the small (10 mL) water samples examined. Ha¨nninen et al. [30] proposed that the volume used for detection of suspected pathogens in drinking water should be at least 8–10 L. Groundwater is considered widely to be microbiologically clean, and is usually used for drinking without treatment. However, groundwater can be contaminated by the environment [26,32]. Stanley et al. [33] showed that Campylobacter can occur in groundwater, with hydrological evidence suggesting that the source of contamination in their study was a dairy farm situated within the hydrological catchment area of the groundwater, which was confirmed when identical strains of C. jejuni were isolated from groundwater and the dairy herd. In the present study, Campylobacter was not identified from groundwater, and dairy herds and other domestic animals in farms around the source were not investigated. However, results of bacterial, chemical and physical water analyses suggested strongly that a river contaminated by agricultural run-off from surrounding sheep and cattle grazing meadows was contaminating the source. In addition, the immediate area of the source was insufficiently protected and vulnerable to the environment. Wild birds, rodents, pigs and cows are all hosts for C. coli and could contaminate surface water easily [26,27,34,35]. Campylobacters are sensitive to many environmental factors and to chlorine [33,36,37], but a failure of the chlorination system could have resulted in the presence of a high concentration of C. coli. Waterborne outbreaks are often caused by several factors that act together [18,20,23,38]. In some outbreaks, heavy rainfall has been responsible for increasing agricultural run-off into rivers, although this did not occur before the outbreak in Gourdon. This underlies the need for increased frequency of routine water quality monitoring, chlorination of the water supply according to environmental conditions, and implementation of preventive measures. This outbreak raised the question of the delay in raising the alert. An outbreak can be recognised by: (i) the poor quality of water detected during regular controls; (ii) complaints about water quality; (iii) an incident in the supply and ℠or sewage water system during maintenance work; and (iv) an increased number of cases of gastroenteritis. In most instances, waterborne outbreaks are suspected when there are already a large number of declared cases [3,12,18,28,38,39]. In the present outbreak, the initial warning was given by a GP who notified officials 5–10 days after the assumed date of onset. This highlights the role of GPs in the detection of an unusual event, particularly during the summer, when an increase in acute gastroenteritis is less expected than during winter [40,41]. In addition, a routine water quality monitoring system able to detect contamination of drinking water before the occurrence of illness and complaints about the water quality should be used as an early warning source. The retrospective cohort study of Gourdon residents allowed the scale and severity of the outbreak to be quantified precisely and the hypothesis that tap water was the vehicle of the outbreak to be tested. When provided with a cohort sample representative of the study population, the number of people affected can be estimated. However, as the exposure of the population was high, a relatively large sample size was needed to test the association between the disease and the exposure. Because of misclassification of exposure (unrecognised consumption of tap water), the attributable risk is probably underestimated. In the absence of obvious alternative explanations, water contamination is the most likely common cause of an explosive outbreak of AGE occurring in a large population. Tillett et al. [42] proposed categorising levels of evidence for waterborne outbreaks. A descriptive study suggesting a water-related outbreak, together with the identification of the same pathogen in stools and water, should be sufficient to conclude that an outbreak is associated strongly with water consumption. Faecal indicators of water quality are currently insufficiently sensitive for some pathogens, notably viruses and parasites (e.g., Cryptosporidium). Therefore, in the absence of evidence from an analytical study to demonstrate an association, the link between illness and water can only be classified as probable or possible. However, total coliform counts are good indicators of faecal contamination with other pathogens (e.g., rotaviruses, E. coli, Campylobacter) [43]. In conclusion, the waterborne outbreak described in the present study was most probably caused by faecal contamination of source A by groundwater, resulting from deficiencies in the maintenance of the water distribution system. Corrective measures were implemented based on the results of the investigation. To our knowledge, this is the first documented waterborne outbreak involving human C. coli infection. In-depth environmental investigations to identify the origin of water contamination are important in order to better understand microbial transmission from the environment and to implement appropriate controls and preventive measures. In addition, surveys on vehicles and vectors that transmit Campylobacter between hosts are crucial in order to better understand the epidemiology of these organisms.

How many people were dead?

297

A large multi-pathogen waterborne community outbreak linked to faecal contamination of a groundwater system, France, 2000

A large waterborne outbreak of infection that occurred during August 2000 in a local community in France was investigated initially via a rapid survey of visits to local physicians. A retrospective cohort study was then conducted on a random cluster sample of residents. Of 709 residents interviewed, 202 (28.5%) were definite cases (at least three liquid stools ℠day or vomiting) and 62 (8.7%) were probable cases (less than three liquid stools ℠day or abdominal pain). Those who had drunk tap water had a threefold increased risk for illness (95% CI 2.4–4.0). The risk increased with the amount of water consumed (chi-square trend: p < 0.0001). Bacteriological analyses of stools were performed for 35 patients and virological analyses for 24 patients. Campylobacter coli, group A rotavirus and norovirus were detected in 31.5%, 71.0% and 21% of samples, respectively. An extensive environmental investigation concluded that a groundwater source to this community had probably been contaminated by agricultural run-off, and a failure in the chlorination system was identified. This is the first documented waterborne outbreak of infection involving human C. coli infections. A better understanding of the factors influencing campylobacter transmission between hosts is required. Keywords Campylobacter coli, epidemiology, gastroenteritis, norovirus, rotavirus, waterborne outbreak INTRODUCTION Contaminated drinking water causes extensive outbreaks of illness because of the large number of people served by water supply facilities. Detection of such outbreaks requires the identification of an increase in illness (usually the rate of gastrointestinal disease) in the exposed population and confirmation that water was the route of transmission [1]. Many countries have routine reporting systems for detecting foodborne and waterborne outbreaks of infection. However, in routine practice it is not easy to detect an increase in clinical cases and to link this increase to waterborne transmission; thus waterborne outbreaks are often unrecognised and underestimated [2]. Microbiological examination of water is more complex than examination of stools, and there is often a failure to detect pathogens in water. Norovirus, Cryptosporidium parvum, Giardia intestinalis and Campylobacter jejuni are the pathogens identified most frequently [3]. The present study describes a large waterborne community outbreak of infection with multiple pathogens, including Campylobacter coli. METHODS AND MATERIALS Background On 23 August 2000, a general practitioner (GP) informed the local district health department that 16 cases of acute gastroenteritis (AGE) had occurred among residents of a holiday camp in the Gourdon community since 21 August. Gourdon, which is located in the south-west of France, has a population of 4888 inhabitants, which increases by 50% in the July– August tourist season. Investigations at the holiday camp concluded that foodborne transmission was very unlikely, but found that tap water was highly contaminated with faecal coliforms. Preliminary information collected from physicians in Gourdon indicated an increase of the number of consultations for AGE. Gourdon tap water is supplied by two underground water facilities: facility A serving 3800 water consumers in Gourdon; and facility B serving 1200 water consumers in Gourdon and the surrounding communities. Water treatment in facility A consists of chlorination, which involves two automatic pumps working in rotation. The water then passes through two semi in-ground reservoir tanks (1100 m3 and 200 m3, respectively) before distribution. Facility B water is mixed with a fraction of facility A water in another 500 m3 semi in-ground reservoir tank before being distributed. This mixing with the treated water from facility A is intended to disinfect the water from facility B. Since the increase in AGE was noted only in Gourdon, contaminated tap water from facility A was suspected to be the origin of the outbreak. Therefore, the population of Gourdon was informed on 25 August to avoid the consumption of tap water or to boil it for 5 min. Survey of gastroenteritis-related medical consultations To confirm the outbreak, the 11 GPs in Gourdon and the local hospital emergency department were requested to provide the number of total daily visits and the number of cases of gastroenteritis cases occurring between 1 August and 11 September 2000. Retrospective cohort study A retrospective cohort study was performed on a sample of Gourdon households between 5 and 15 September 2000. Domestic households were selected randomly from the telephone directory. A definite case was defined as a Gourdon resident who had at least three liquid stools ℠day or vomiting between 1 August and 3 September, and who had been present in Gourdon for ‡ 1 day between 1 and 31 August; if less than three liquid stools ℠day or abdominal pain had occurred, the case was classified as probable. Exposure was defined as the consumption of tap water before 25 August 2000. A standardised telephone questionnaire was completed to obtain demographic data (gender, date of birth), food consumption (commercially prepared dishes, sea-food, raw milk, pastry), tap water consumption (before and after 25 August, daily number of glasses of water, consumption of boiled tap water), symptoms of gastroenteritis (date of onset, number of liquid stools ℠day, vomiting, nausea, abdominal pain, fever), medical care (in- or out-patient) and consequences on daily activities (sick leave, bed confinement) for each member of the selected household. Under the hypothesis that 60% of the population consumed tap water, an estimated 550 people were required for inclusion in the study to identify a relative risk (RR) of ‡ 3, with a 5% a risk and a power of 80%. Based on an average of two individuals per household (INSERM 1999 population census), 300 households were required for inclusion in the survey. The association between AGE and the consumption of tap water was assessed by calculating the risk ratio (RR) and 95% CI, taking into account a household cluster design effect. A dose-effect relationship was evaluated by analysing the RR trend for different categories of daily number of glasses of tap water consumed (chi-square trend). Proportions were compared using the corrected chi-square test, and means were compared using the Student test or Mann and Witney test. Data were analysed with Epi-Info 6.04c (CDC, Atlanta, GA, USA). The gender and age distribution of the cohort population were compared with that of the general population of Gourdon (INSERM 1999 population census) to assess whether it was representative. Under the hypothesis of representation, the total number of cases of gastroenteritis was estimated by applying the attack rate to the total population of Gourdon, including the tourist population (data provided by the Gourdon tourism office). The most likely period for contamination of the water distribution system was based on the epidemic curve and the minimal incubation period of the pathogens involved. Microbiological analyses of stools Stool samples were examined for the presence of Salmonella spp., Shigella spp., Staphylococcus aureus, Campylobacter spp., Yersinia enterocolitica, Escherichia coli, enterococci, rotavirus groups A and C, astrovirus, calicivirus, adenovirus types 40 and 41, enterovirus, hepatitis A virus and Cryptosporidum. Viruses were detected using an enzyme immunoassay for group A rotavirus, astrovirus and adenovirus 40 ℠41, and by RT-PCR for calicivirus, group C rotavirus, enterovirus and hepatitis A virus [4–6]. Campylobacter isolates were sent to the National Reference Centre for Campylobacter for identification and fingerprinting. Species identification was by means of phenotypic tests (API Campy; bioMe´rieux, Lyon, France), with use of PCR to differentiate Campylobacter jejuni from Camplylobacter coli [7]. Random amplified polymorphic DNA fingerprinting was performed with primer 3881, 5¢-AACGCGCAAC. Environmental investigation An environmental and microbiological investigation of the water distribution system was performed. The inspection concerned the water catchments, the water distribution and sewage system, agricultural practices and private households around the sources. Bacteriological (Escherichia coli, enterococci, faecal streptococci, thermotolerant coliforms, sulphite-reducing clostridia) and chemical (chlorine) analyses were performed on tap water and on the water sampled from the two groundwater sources. After 23 August, the water chlorine concentration and the presence of coliform bacteria were monitored daily at several sites of the water system before and after the resort treatment. Examinations for viruses (group A rotavirus, astrovirus, norovirus, enterovirus, hepatitis A virus) and Campylobacter spp. were undertaken on two 10-mL tap water samples collected inhouse on 23 August, and on two 10-mL samples from groundwater sources A and B on 25 August. Viruses were detected by RT-PCR, followed by hybridisation with a specific primer to improve sensitivity, while C. jejuni and C. coli were detected by PCR with specific primers. Group A rotavirus isolates detected in stools and in water were sequenced using Beg 9 and End 9 primers [8] and compared using the Infobiogene GCG software (http://www.infobiogen.fr). Bacteriological (revived 36C, revived 22C, thermotolerant coliforms, streptococci and faecal enterococci), physical (temperature, conductivity) and chemical (oxygen consumption, ammonium, nitrates, phosphorus) parameters from source A and the river were analysed. Trends were compared for an 8-week period between 21 December 2000 and 7 February 2001. RESULTS Epidemiology Between 1 August and 11 September 2000, 7104 Gourdon residents consulted a GP or the local hospital emergency department. Between 24 and 28 August, gastroenteritis accounted for 44% (479 ℠1097) of all medical visits, compared with 6% (126 ℠2185) during the first 15 days of August (Fig. 1). Among the 498 households contacted for the household survey, 198 were excluded (103 did not respond, 49 did not fulfil the inclusion criteria, 43 refused to participate, and three for unknown reasons). The questionnaire was completed for 300 households (709 individuals). Data were missing concerning gender for one individual and age for 11 individuals. Overall, 331 (46.7%; 95% CI 44.5–49.0) of respondents were male and 377 (53.2%; 95% CI 50.9–55.5) were female. The median age was 44 years (range < 1–94 years). The age and gender distribution of the sample was similar to that of the general population. During the outbreak period, 264 of the 709 respondents (attack rate (AR) 37.2%) had been ill, with 202 definite cases (AR 28.5%; 95% CI 24.6– 32.4) and 62 probable cases (AR 8.7%; 95% CI 6.5– 11.0). The AR was higher in females (31.8%) than in males (24.8%), and was highest in children aged < 6 years (42.1%, Table 1). The epidemic curve (Fig. 2) suggests that the outbreak started between 14 and 19 August. The shape of the epidemic curve in the retrospective cohort study was very similar to that based on the number of visits to GPs for gastroenteritis. Patients complained mainly of diarrhoea, abdominal pain and nausea (Table 2). No cases of bloody diarrhoea were notified. The mean duration of symptoms was 4.5 days (median 3, range 1–30 days). Among definite cases, 52% had visited a GP, 42% had taken sick leave for ‡ 1 day, 32% had required bed rest, and 2.9% (six patients) had been hospitalised. In total, 336 (47.4%) respondents had drunk tap water before 25 August, whereas 646 (91.1%) did not drink tap water after 25 August. The risk for illness, taking into account the sample cluster design effect of 1.4, was three-fold greater (95% CI 2.2–4.1) for those who had drunk tap water before 25 August, and increased with the amount of water consumed (chi-square trend: p < 0.0001; Table 3). The risk of gastroenteritis associated with drinking tap water did not change substantially after adjustment for age (RRMH 3.3, 95% CI 2.5–4.4), and there was no modification effect according to age. Food consumption was not associated with illness. Applying the AR of 37% to the total population of Gourdon exposed to the water supply (4888 residents + 2200 tourists), the number of individuals affected by this outbreak was estimated to be 2600 (95% CI 2400–2900), assuming an exposure to tap water and a risk for gastroenteritis among the tourists similar to that among the inhabitants. Microbiological stool analyses Among the 35 stool samples analysed for bacteria, 11 (31.5%) were positive for C. coli. Twenty-four of the 35 samples were also tested for viruses: rotavirus A was detected in 17 (71.0%) and norovirus in five (21.0%) samples, while nine (37.5%) samples were co-infected (four C. coli + rotavirus; one C. coli + norovirus; three rotavirus + norovirus; and one C. coli + rotavirus + norovirus). Of the 11 C. coli isolates, six were characterised as biotype 2, and were identical following RAPD analysis, while the remaining five were of biotype 1. Rotavirus typing by RTPCR identified a single genotype P-(8), G1. Molecular characterisation of the norovirus isolates identified two different patterns, with three isolates being identical and close to the Saragota strain (genogroup I) and two close to the new GGIIb variant (genogroup II). Environmental investigation All of the water samples from source B were negative, regardless of the site of the distribution system tested. Analyses of the water originating from source A, sampled after leaving the water-works on 23 and 25 August, revealed high concentrations of E. coli (2263 / 100 mL), enterococci (415 / 100 mL), total coliforms (90 / 100 mL), faecal streptococci (136 / 100 mL) and sulphite-reducing clostridia (10 / 100 mL), and an absence of chlorination. Rotavirus A genotype G1 was detected in the crude water sampled from source A. The analyses for enterovirus, astrovirus, norovirus, hepatitis A virus and Campylobacter were negative. Source A was situated in a valley containing a river, below a hamlet of 15 dwellings, including farms breeding cattle, pork and sheep. The source was not supposed to be supplied by the river. The river resulted from the convergence of about ten different streams, with one originating close to a slaughterhouse, and was exposed to agricultural run-off and manure effluent; sheep and cattle grazed in meadows bordering the river, and a large amount of sewage from water treatment facilities was sprayed on to land close to the river. Cowsheds were close to the water catchment area. The immediate protected area around source A was too small and littered with building rubbish. Air vents without protection in the door made the source accessible to animals and insects. Bacteriological analyses of the water samples from source A and the river revealed that the contamination was of both human and animal origin, with a higher concentration in the river. The levels and trends of the chemical parameters of source A were similar to those of the river, strongly suggesting that the river was connected to source A (Fig. 3). A retrospective review of routine data collected between 1991 and 2000 showed irregularities in the maintenance of the water distribution system and the water treatment plant, particularly during the month of August. In the hamlet, the sewage system of the 15 households did not comply with sanitary rules. All but three ejected waste either into a cesspool or superficially without treatment. Three regions of gene 9 of the rotavirus detected in stools and water were compared. Sequencing showed that the strains were closely related but not identical. Assuming a likely onset of the outbreak between 14 and 19 August, and taking into account a minimum incubation period of 2 days for rotavirus A, norovirus and C. coli, the contamination of the water distribution system probably occurred between 12 and 17 August. Contamination probably persisted until 28 August, as intensive chlorination, beginning on 25 August, was effective only from 28 August onwards (Fig. 2). DISCUSSION The present descriptive and retrospective cohort studies confirmed that the outbreak of infection in the community of Gourdon was related to the consumption of tap water and affected 2600 individuals. The outbreak involved multiple pathogens, with group A rotavirus and C. coli being the most frequent, followed by norovirus. Results of environmental investigations suggested that water contamination was related to the contamination of the source, combined with a failure of the water chlorination system. The association with tap water consumption, the dose-effect relationship, and the fact that multiple pathogens were involved, were all strongly in favour of tap water as the origin of infection. Moreover, the number of cases decreased dramatically after the water restriction advice and the implementation of control measures on the water supply. No food products explored in the study were associated with illness. An AR of 14.5% and 31% was observed among adults and children aged < 6 years, respectively, who reported that they had not consumed tap water. This relatively high rate could be explained by several phenomena. Many individuals may have consumed tap water without being aware of it, e.g., in drinks or ice cubes consumed outside of the home, via raw fruits and vegetables, dishes prepared with unboiled water, or while cleaning their teeth. It can be hypothesised that people who became ill without being aware of having consumed tap water were exposed to small amounts of water and had been subjected to a small infective dose. In addition, children, more than adults, may have been contaminated through person-to-person transmission (e.g., family contact), which is the most common route of transmission for rotavirus and calicivirus. Waterborne contamination can affect large numbers of individuals [3,9–11]. In the present study, assuming a 50% increase in the population during the tourist season, the number of affected people was estimated to be between 2400 and 2900. The cohort study showed that almost half (46.5%) of the cases consulted a GP. Applying this proportion to the estimated number of cases yielded an estimated 1116–1349 ill individuals who consulted a GP for gastroenteritis, similar to the number of consultations for gastroenteritis actually reported by the GPs (n = 1037). The stool sample analyses revealed the presence of multiple pathogens (rotavirus A, norovirus genogroups I and II and C. coli) as single or co-infections. Except for C. coli, these pathogens have been implicated in numerous waterborne outbreaks, suggesting an influx of sewage into the water supply system [10,12–18]. Rotaviruses are known to be responsible for large epidemics during winter seasons, affecting particularly children aged < 3 years. These epidemics are caused by the faecal route of contamination (personto-person transmission). The high proportion of stools positive for rotavirus (71%) is uncommon in waterborne outbreaks and probably reflects massive contamination [15,16,19]. Different molecular sequences of rotavirus A genotype G1 found in stools and water indicated a human faecal source of infection. In the present study, contamination of water by sewage could be linked with the large amount of sewage from the water purification plant sprayed on to land close to the river and the water catchments. In this outbreak, C. coli, which was identified in 32% of stools, was the second most frequent pathogen. To our knowledge, human C. coli contamination has not been described previously as a causative agent in waterborne outbreaks, unlike C. jejuni [11,20–25], although C. coli has been isolated from water [26]. The main symptoms observed (diarrhoea, abdominal pain, nausea) and the mean duration of the disease (4.5 days) were compatible with a Campylobacter infection. No bloody diarrhoea was reported, although bloody diarrhoea is not constant in Campylobacter infections [27]. The high proportion of individuals who had to take sick leave, and the fact that six individuals were hospitalised, suggested a relative severity of symptoms, which is much more compatible with Campylobacter infection than with virus infections. Among the hospitalised patients, 60% were infected with C. coli. Among patients whose stools were analysed and whose age was known, cases infected only with rotavirus were younger (mean age 49 years; range, 5 months to 95 years) than cases infected only with Campylobacter (mean age 66 years; range 23–88 years) or co-infected with rotavirus plus Campylobacter (mean age 69 years; range 2–91 years). Moreover, the proportion (52%) of medical visits was unusually high compared with the proportion of visits observed during other waterborne outbreaks [13,16,28]. Campylobacter needs specific atmospheric conditions for growth and survival and its detection can be very difficult [29,30]. Because of a rapid physiological transformation to a viable, though non-cultivable, state, it has been suggested that conventional culture methods often fail [31]. Nevertheless, in the present study, the failure to detect C. coli in water, and the detection of noroviruses in stools, could be caused by the small (10 mL) water samples examined. Ha¨nninen et al. [30] proposed that the volume used for detection of suspected pathogens in drinking water should be at least 8–10 L. Groundwater is considered widely to be microbiologically clean, and is usually used for drinking without treatment. However, groundwater can be contaminated by the environment [26,32]. Stanley et al. [33] showed that Campylobacter can occur in groundwater, with hydrological evidence suggesting that the source of contamination in their study was a dairy farm situated within the hydrological catchment area of the groundwater, which was confirmed when identical strains of C. jejuni were isolated from groundwater and the dairy herd. In the present study, Campylobacter was not identified from groundwater, and dairy herds and other domestic animals in farms around the source were not investigated. However, results of bacterial, chemical and physical water analyses suggested strongly that a river contaminated by agricultural run-off from surrounding sheep and cattle grazing meadows was contaminating the source. In addition, the immediate area of the source was insufficiently protected and vulnerable to the environment. Wild birds, rodents, pigs and cows are all hosts for C. coli and could contaminate surface water easily [26,27,34,35]. Campylobacters are sensitive to many environmental factors and to chlorine [33,36,37], but a failure of the chlorination system could have resulted in the presence of a high concentration of C. coli. Waterborne outbreaks are often caused by several factors that act together [18,20,23,38]. In some outbreaks, heavy rainfall has been responsible for increasing agricultural run-off into rivers, although this did not occur before the outbreak in Gourdon. This underlies the need for increased frequency of routine water quality monitoring, chlorination of the water supply according to environmental conditions, and implementation of preventive measures. This outbreak raised the question of the delay in raising the alert. An outbreak can be recognised by: (i) the poor quality of water detected during regular controls; (ii) complaints about water quality; (iii) an incident in the supply and ℠or sewage water system during maintenance work; and (iv) an increased number of cases of gastroenteritis. In most instances, waterborne outbreaks are suspected when there are already a large number of declared cases [3,12,18,28,38,39]. In the present outbreak, the initial warning was given by a GP who notified officials 5–10 days after the assumed date of onset. This highlights the role of GPs in the detection of an unusual event, particularly during the summer, when an increase in acute gastroenteritis is less expected than during winter [40,41]. In addition, a routine water quality monitoring system able to detect contamination of drinking water before the occurrence of illness and complaints about the water quality should be used as an early warning source. The retrospective cohort study of Gourdon residents allowed the scale and severity of the outbreak to be quantified precisely and the hypothesis that tap water was the vehicle of the outbreak to be tested. When provided with a cohort sample representative of the study population, the number of people affected can be estimated. However, as the exposure of the population was high, a relatively large sample size was needed to test the association between the disease and the exposure. Because of misclassification of exposure (unrecognised consumption of tap water), the attributable risk is probably underestimated. In the absence of obvious alternative explanations, water contamination is the most likely common cause of an explosive outbreak of AGE occurring in a large population. Tillett et al. [42] proposed categorising levels of evidence for waterborne outbreaks. A descriptive study suggesting a water-related outbreak, together with the identification of the same pathogen in stools and water, should be sufficient to conclude that an outbreak is associated strongly with water consumption. Faecal indicators of water quality are currently insufficiently sensitive for some pathogens, notably viruses and parasites (e.g., Cryptosporidium). Therefore, in the absence of evidence from an analytical study to demonstrate an association, the link between illness and water can only be classified as probable or possible. However, total coliform counts are good indicators of faecal contamination with other pathogens (e.g., rotaviruses, E. coli, Campylobacter) [43]. In conclusion, the waterborne outbreak described in the present study was most probably caused by faecal contamination of source A by groundwater, resulting from deficiencies in the maintenance of the water distribution system. Corrective measures were implemented based on the results of the investigation. To our knowledge, this is the first documented waterborne outbreak involving human C. coli infection. In-depth environmental investigations to identify the origin of water contamination are important in order to better understand microbial transmission from the environment and to implement appropriate controls and preventive measures. In addition, surveys on vehicles and vectors that transmit Campylobacter between hosts are crucial in order to better understand the epidemiology of these organisms.

Which contaminants or viruses or bacteria were found?

298

A large multi-pathogen waterborne community outbreak linked to faecal contamination of a groundwater system, France, 2000

A large waterborne outbreak of infection that occurred during August 2000 in a local community in France was investigated initially via a rapid survey of visits to local physicians. A retrospective cohort study was then conducted on a random cluster sample of residents. Of 709 residents interviewed, 202 (28.5%) were definite cases (at least three liquid stools ℠day or vomiting) and 62 (8.7%) were probable cases (less than three liquid stools ℠day or abdominal pain). Those who had drunk tap water had a threefold increased risk for illness (95% CI 2.4–4.0). The risk increased with the amount of water consumed (chi-square trend: p < 0.0001). Bacteriological analyses of stools were performed for 35 patients and virological analyses for 24 patients. Campylobacter coli, group A rotavirus and norovirus were detected in 31.5%, 71.0% and 21% of samples, respectively. An extensive environmental investigation concluded that a groundwater source to this community had probably been contaminated by agricultural run-off, and a failure in the chlorination system was identified. This is the first documented waterborne outbreak of infection involving human C. coli infections. A better understanding of the factors influencing campylobacter transmission between hosts is required. Keywords Campylobacter coli, epidemiology, gastroenteritis, norovirus, rotavirus, waterborne outbreak INTRODUCTION Contaminated drinking water causes extensive outbreaks of illness because of the large number of people served by water supply facilities. Detection of such outbreaks requires the identification of an increase in illness (usually the rate of gastrointestinal disease) in the exposed population and confirmation that water was the route of transmission [1]. Many countries have routine reporting systems for detecting foodborne and waterborne outbreaks of infection. However, in routine practice it is not easy to detect an increase in clinical cases and to link this increase to waterborne transmission; thus waterborne outbreaks are often unrecognised and underestimated [2]. Microbiological examination of water is more complex than examination of stools, and there is often a failure to detect pathogens in water. Norovirus, Cryptosporidium parvum, Giardia intestinalis and Campylobacter jejuni are the pathogens identified most frequently [3]. The present study describes a large waterborne community outbreak of infection with multiple pathogens, including Campylobacter coli. METHODS AND MATERIALS Background On 23 August 2000, a general practitioner (GP) informed the local district health department that 16 cases of acute gastroenteritis (AGE) had occurred among residents of a holiday camp in the Gourdon community since 21 August. Gourdon, which is located in the south-west of France, has a population of 4888 inhabitants, which increases by 50% in the July– August tourist season. Investigations at the holiday camp concluded that foodborne transmission was very unlikely, but found that tap water was highly contaminated with faecal coliforms. Preliminary information collected from physicians in Gourdon indicated an increase of the number of consultations for AGE. Gourdon tap water is supplied by two underground water facilities: facility A serving 3800 water consumers in Gourdon; and facility B serving 1200 water consumers in Gourdon and the surrounding communities. Water treatment in facility A consists of chlorination, which involves two automatic pumps working in rotation. The water then passes through two semi in-ground reservoir tanks (1100 m3 and 200 m3, respectively) before distribution. Facility B water is mixed with a fraction of facility A water in another 500 m3 semi in-ground reservoir tank before being distributed. This mixing with the treated water from facility A is intended to disinfect the water from facility B. Since the increase in AGE was noted only in Gourdon, contaminated tap water from facility A was suspected to be the origin of the outbreak. Therefore, the population of Gourdon was informed on 25 August to avoid the consumption of tap water or to boil it for 5 min. Survey of gastroenteritis-related medical consultations To confirm the outbreak, the 11 GPs in Gourdon and the local hospital emergency department were requested to provide the number of total daily visits and the number of cases of gastroenteritis cases occurring between 1 August and 11 September 2000. Retrospective cohort study A retrospective cohort study was performed on a sample of Gourdon households between 5 and 15 September 2000. Domestic households were selected randomly from the telephone directory. A definite case was defined as a Gourdon resident who had at least three liquid stools ℠day or vomiting between 1 August and 3 September, and who had been present in Gourdon for ‡ 1 day between 1 and 31 August; if less than three liquid stools ℠day or abdominal pain had occurred, the case was classified as probable. Exposure was defined as the consumption of tap water before 25 August 2000. A standardised telephone questionnaire was completed to obtain demographic data (gender, date of birth), food consumption (commercially prepared dishes, sea-food, raw milk, pastry), tap water consumption (before and after 25 August, daily number of glasses of water, consumption of boiled tap water), symptoms of gastroenteritis (date of onset, number of liquid stools ℠day, vomiting, nausea, abdominal pain, fever), medical care (in- or out-patient) and consequences on daily activities (sick leave, bed confinement) for each member of the selected household. Under the hypothesis that 60% of the population consumed tap water, an estimated 550 people were required for inclusion in the study to identify a relative risk (RR) of ‡ 3, with a 5% a risk and a power of 80%. Based on an average of two individuals per household (INSERM 1999 population census), 300 households were required for inclusion in the survey. The association between AGE and the consumption of tap water was assessed by calculating the risk ratio (RR) and 95% CI, taking into account a household cluster design effect. A dose-effect relationship was evaluated by analysing the RR trend for different categories of daily number of glasses of tap water consumed (chi-square trend). Proportions were compared using the corrected chi-square test, and means were compared using the Student test or Mann and Witney test. Data were analysed with Epi-Info 6.04c (CDC, Atlanta, GA, USA). The gender and age distribution of the cohort population were compared with that of the general population of Gourdon (INSERM 1999 population census) to assess whether it was representative. Under the hypothesis of representation, the total number of cases of gastroenteritis was estimated by applying the attack rate to the total population of Gourdon, including the tourist population (data provided by the Gourdon tourism office). The most likely period for contamination of the water distribution system was based on the epidemic curve and the minimal incubation period of the pathogens involved. Microbiological analyses of stools Stool samples were examined for the presence of Salmonella spp., Shigella spp., Staphylococcus aureus, Campylobacter spp., Yersinia enterocolitica, Escherichia coli, enterococci, rotavirus groups A and C, astrovirus, calicivirus, adenovirus types 40 and 41, enterovirus, hepatitis A virus and Cryptosporidum. Viruses were detected using an enzyme immunoassay for group A rotavirus, astrovirus and adenovirus 40 ℠41, and by RT-PCR for calicivirus, group C rotavirus, enterovirus and hepatitis A virus [4–6]. Campylobacter isolates were sent to the National Reference Centre for Campylobacter for identification and fingerprinting. Species identification was by means of phenotypic tests (API Campy; bioMe´rieux, Lyon, France), with use of PCR to differentiate Campylobacter jejuni from Camplylobacter coli [7]. Random amplified polymorphic DNA fingerprinting was performed with primer 3881, 5¢-AACGCGCAAC. Environmental investigation An environmental and microbiological investigation of the water distribution system was performed. The inspection concerned the water catchments, the water distribution and sewage system, agricultural practices and private households around the sources. Bacteriological (Escherichia coli, enterococci, faecal streptococci, thermotolerant coliforms, sulphite-reducing clostridia) and chemical (chlorine) analyses were performed on tap water and on the water sampled from the two groundwater sources. After 23 August, the water chlorine concentration and the presence of coliform bacteria were monitored daily at several sites of the water system before and after the resort treatment. Examinations for viruses (group A rotavirus, astrovirus, norovirus, enterovirus, hepatitis A virus) and Campylobacter spp. were undertaken on two 10-mL tap water samples collected inhouse on 23 August, and on two 10-mL samples from groundwater sources A and B on 25 August. Viruses were detected by RT-PCR, followed by hybridisation with a specific primer to improve sensitivity, while C. jejuni and C. coli were detected by PCR with specific primers. Group A rotavirus isolates detected in stools and in water were sequenced using Beg 9 and End 9 primers [8] and compared using the Infobiogene GCG software (http://www.infobiogen.fr). Bacteriological (revived 36C, revived 22C, thermotolerant coliforms, streptococci and faecal enterococci), physical (temperature, conductivity) and chemical (oxygen consumption, ammonium, nitrates, phosphorus) parameters from source A and the river were analysed. Trends were compared for an 8-week period between 21 December 2000 and 7 February 2001. RESULTS Epidemiology Between 1 August and 11 September 2000, 7104 Gourdon residents consulted a GP or the local hospital emergency department. Between 24 and 28 August, gastroenteritis accounted for 44% (479 ℠1097) of all medical visits, compared with 6% (126 ℠2185) during the first 15 days of August (Fig. 1). Among the 498 households contacted for the household survey, 198 were excluded (103 did not respond, 49 did not fulfil the inclusion criteria, 43 refused to participate, and three for unknown reasons). The questionnaire was completed for 300 households (709 individuals). Data were missing concerning gender for one individual and age for 11 individuals. Overall, 331 (46.7%; 95% CI 44.5–49.0) of respondents were male and 377 (53.2%; 95% CI 50.9–55.5) were female. The median age was 44 years (range < 1–94 years). The age and gender distribution of the sample was similar to that of the general population. During the outbreak period, 264 of the 709 respondents (attack rate (AR) 37.2%) had been ill, with 202 definite cases (AR 28.5%; 95% CI 24.6– 32.4) and 62 probable cases (AR 8.7%; 95% CI 6.5– 11.0). The AR was higher in females (31.8%) than in males (24.8%), and was highest in children aged < 6 years (42.1%, Table 1). The epidemic curve (Fig. 2) suggests that the outbreak started between 14 and 19 August. The shape of the epidemic curve in the retrospective cohort study was very similar to that based on the number of visits to GPs for gastroenteritis. Patients complained mainly of diarrhoea, abdominal pain and nausea (Table 2). No cases of bloody diarrhoea were notified. The mean duration of symptoms was 4.5 days (median 3, range 1–30 days). Among definite cases, 52% had visited a GP, 42% had taken sick leave for ‡ 1 day, 32% had required bed rest, and 2.9% (six patients) had been hospitalised. In total, 336 (47.4%) respondents had drunk tap water before 25 August, whereas 646 (91.1%) did not drink tap water after 25 August. The risk for illness, taking into account the sample cluster design effect of 1.4, was three-fold greater (95% CI 2.2–4.1) for those who had drunk tap water before 25 August, and increased with the amount of water consumed (chi-square trend: p < 0.0001; Table 3). The risk of gastroenteritis associated with drinking tap water did not change substantially after adjustment for age (RRMH 3.3, 95% CI 2.5–4.4), and there was no modification effect according to age. Food consumption was not associated with illness. Applying the AR of 37% to the total population of Gourdon exposed to the water supply (4888 residents + 2200 tourists), the number of individuals affected by this outbreak was estimated to be 2600 (95% CI 2400–2900), assuming an exposure to tap water and a risk for gastroenteritis among the tourists similar to that among the inhabitants. Microbiological stool analyses Among the 35 stool samples analysed for bacteria, 11 (31.5%) were positive for C. coli. Twenty-four of the 35 samples were also tested for viruses: rotavirus A was detected in 17 (71.0%) and norovirus in five (21.0%) samples, while nine (37.5%) samples were co-infected (four C. coli + rotavirus; one C. coli + norovirus; three rotavirus + norovirus; and one C. coli + rotavirus + norovirus). Of the 11 C. coli isolates, six were characterised as biotype 2, and were identical following RAPD analysis, while the remaining five were of biotype 1. Rotavirus typing by RTPCR identified a single genotype P-(8), G1. Molecular characterisation of the norovirus isolates identified two different patterns, with three isolates being identical and close to the Saragota strain (genogroup I) and two close to the new GGIIb variant (genogroup II). Environmental investigation All of the water samples from source B were negative, regardless of the site of the distribution system tested. Analyses of the water originating from source A, sampled after leaving the water-works on 23 and 25 August, revealed high concentrations of E. coli (2263 / 100 mL), enterococci (415 / 100 mL), total coliforms (90 / 100 mL), faecal streptococci (136 / 100 mL) and sulphite-reducing clostridia (10 / 100 mL), and an absence of chlorination. Rotavirus A genotype G1 was detected in the crude water sampled from source A. The analyses for enterovirus, astrovirus, norovirus, hepatitis A virus and Campylobacter were negative. Source A was situated in a valley containing a river, below a hamlet of 15 dwellings, including farms breeding cattle, pork and sheep. The source was not supposed to be supplied by the river. The river resulted from the convergence of about ten different streams, with one originating close to a slaughterhouse, and was exposed to agricultural run-off and manure effluent; sheep and cattle grazed in meadows bordering the river, and a large amount of sewage from water treatment facilities was sprayed on to land close to the river. Cowsheds were close to the water catchment area. The immediate protected area around source A was too small and littered with building rubbish. Air vents without protection in the door made the source accessible to animals and insects. Bacteriological analyses of the water samples from source A and the river revealed that the contamination was of both human and animal origin, with a higher concentration in the river. The levels and trends of the chemical parameters of source A were similar to those of the river, strongly suggesting that the river was connected to source A (Fig. 3). A retrospective review of routine data collected between 1991 and 2000 showed irregularities in the maintenance of the water distribution system and the water treatment plant, particularly during the month of August. In the hamlet, the sewage system of the 15 households did not comply with sanitary rules. All but three ejected waste either into a cesspool or superficially without treatment. Three regions of gene 9 of the rotavirus detected in stools and water were compared. Sequencing showed that the strains were closely related but not identical. Assuming a likely onset of the outbreak between 14 and 19 August, and taking into account a minimum incubation period of 2 days for rotavirus A, norovirus and C. coli, the contamination of the water distribution system probably occurred between 12 and 17 August. Contamination probably persisted until 28 August, as intensive chlorination, beginning on 25 August, was effective only from 28 August onwards (Fig. 2). DISCUSSION The present descriptive and retrospective cohort studies confirmed that the outbreak of infection in the community of Gourdon was related to the consumption of tap water and affected 2600 individuals. The outbreak involved multiple pathogens, with group A rotavirus and C. coli being the most frequent, followed by norovirus. Results of environmental investigations suggested that water contamination was related to the contamination of the source, combined with a failure of the water chlorination system. The association with tap water consumption, the dose-effect relationship, and the fact that multiple pathogens were involved, were all strongly in favour of tap water as the origin of infection. Moreover, the number of cases decreased dramatically after the water restriction advice and the implementation of control measures on the water supply. No food products explored in the study were associated with illness. An AR of 14.5% and 31% was observed among adults and children aged < 6 years, respectively, who reported that they had not consumed tap water. This relatively high rate could be explained by several phenomena. Many individuals may have consumed tap water without being aware of it, e.g., in drinks or ice cubes consumed outside of the home, via raw fruits and vegetables, dishes prepared with unboiled water, or while cleaning their teeth. It can be hypothesised that people who became ill without being aware of having consumed tap water were exposed to small amounts of water and had been subjected to a small infective dose. In addition, children, more than adults, may have been contaminated through person-to-person transmission (e.g., family contact), which is the most common route of transmission for rotavirus and calicivirus. Waterborne contamination can affect large numbers of individuals [3,9–11]. In the present study, assuming a 50% increase in the population during the tourist season, the number of affected people was estimated to be between 2400 and 2900. The cohort study showed that almost half (46.5%) of the cases consulted a GP. Applying this proportion to the estimated number of cases yielded an estimated 1116–1349 ill individuals who consulted a GP for gastroenteritis, similar to the number of consultations for gastroenteritis actually reported by the GPs (n = 1037). The stool sample analyses revealed the presence of multiple pathogens (rotavirus A, norovirus genogroups I and II and C. coli) as single or co-infections. Except for C. coli, these pathogens have been implicated in numerous waterborne outbreaks, suggesting an influx of sewage into the water supply system [10,12–18]. Rotaviruses are known to be responsible for large epidemics during winter seasons, affecting particularly children aged < 3 years. These epidemics are caused by the faecal route of contamination (personto-person transmission). The high proportion of stools positive for rotavirus (71%) is uncommon in waterborne outbreaks and probably reflects massive contamination [15,16,19]. Different molecular sequences of rotavirus A genotype G1 found in stools and water indicated a human faecal source of infection. In the present study, contamination of water by sewage could be linked with the large amount of sewage from the water purification plant sprayed on to land close to the river and the water catchments. In this outbreak, C. coli, which was identified in 32% of stools, was the second most frequent pathogen. To our knowledge, human C. coli contamination has not been described previously as a causative agent in waterborne outbreaks, unlike C. jejuni [11,20–25], although C. coli has been isolated from water [26]. The main symptoms observed (diarrhoea, abdominal pain, nausea) and the mean duration of the disease (4.5 days) were compatible with a Campylobacter infection. No bloody diarrhoea was reported, although bloody diarrhoea is not constant in Campylobacter infections [27]. The high proportion of individuals who had to take sick leave, and the fact that six individuals were hospitalised, suggested a relative severity of symptoms, which is much more compatible with Campylobacter infection than with virus infections. Among the hospitalised patients, 60% were infected with C. coli. Among patients whose stools were analysed and whose age was known, cases infected only with rotavirus were younger (mean age 49 years; range, 5 months to 95 years) than cases infected only with Campylobacter (mean age 66 years; range 23–88 years) or co-infected with rotavirus plus Campylobacter (mean age 69 years; range 2–91 years). Moreover, the proportion (52%) of medical visits was unusually high compared with the proportion of visits observed during other waterborne outbreaks [13,16,28]. Campylobacter needs specific atmospheric conditions for growth and survival and its detection can be very difficult [29,30]. Because of a rapid physiological transformation to a viable, though non-cultivable, state, it has been suggested that conventional culture methods often fail [31]. Nevertheless, in the present study, the failure to detect C. coli in water, and the detection of noroviruses in stools, could be caused by the small (10 mL) water samples examined. Ha¨nninen et al. [30] proposed that the volume used for detection of suspected pathogens in drinking water should be at least 8–10 L. Groundwater is considered widely to be microbiologically clean, and is usually used for drinking without treatment. However, groundwater can be contaminated by the environment [26,32]. Stanley et al. [33] showed that Campylobacter can occur in groundwater, with hydrological evidence suggesting that the source of contamination in their study was a dairy farm situated within the hydrological catchment area of the groundwater, which was confirmed when identical strains of C. jejuni were isolated from groundwater and the dairy herd. In the present study, Campylobacter was not identified from groundwater, and dairy herds and other domestic animals in farms around the source were not investigated. However, results of bacterial, chemical and physical water analyses suggested strongly that a river contaminated by agricultural run-off from surrounding sheep and cattle grazing meadows was contaminating the source. In addition, the immediate area of the source was insufficiently protected and vulnerable to the environment. Wild birds, rodents, pigs and cows are all hosts for C. coli and could contaminate surface water easily [26,27,34,35]. Campylobacters are sensitive to many environmental factors and to chlorine [33,36,37], but a failure of the chlorination system could have resulted in the presence of a high concentration of C. coli. Waterborne outbreaks are often caused by several factors that act together [18,20,23,38]. In some outbreaks, heavy rainfall has been responsible for increasing agricultural run-off into rivers, although this did not occur before the outbreak in Gourdon. This underlies the need for increased frequency of routine water quality monitoring, chlorination of the water supply according to environmental conditions, and implementation of preventive measures. This outbreak raised the question of the delay in raising the alert. An outbreak can be recognised by: (i) the poor quality of water detected during regular controls; (ii) complaints about water quality; (iii) an incident in the supply and ℠or sewage water system during maintenance work; and (iv) an increased number of cases of gastroenteritis. In most instances, waterborne outbreaks are suspected when there are already a large number of declared cases [3,12,18,28,38,39]. In the present outbreak, the initial warning was given by a GP who notified officials 5–10 days after the assumed date of onset. This highlights the role of GPs in the detection of an unusual event, particularly during the summer, when an increase in acute gastroenteritis is less expected than during winter [40,41]. In addition, a routine water quality monitoring system able to detect contamination of drinking water before the occurrence of illness and complaints about the water quality should be used as an early warning source. The retrospective cohort study of Gourdon residents allowed the scale and severity of the outbreak to be quantified precisely and the hypothesis that tap water was the vehicle of the outbreak to be tested. When provided with a cohort sample representative of the study population, the number of people affected can be estimated. However, as the exposure of the population was high, a relatively large sample size was needed to test the association between the disease and the exposure. Because of misclassification of exposure (unrecognised consumption of tap water), the attributable risk is probably underestimated. In the absence of obvious alternative explanations, water contamination is the most likely common cause of an explosive outbreak of AGE occurring in a large population. Tillett et al. [42] proposed categorising levels of evidence for waterborne outbreaks. A descriptive study suggesting a water-related outbreak, together with the identification of the same pathogen in stools and water, should be sufficient to conclude that an outbreak is associated strongly with water consumption. Faecal indicators of water quality are currently insufficiently sensitive for some pathogens, notably viruses and parasites (e.g., Cryptosporidium). Therefore, in the absence of evidence from an analytical study to demonstrate an association, the link between illness and water can only be classified as probable or possible. However, total coliform counts are good indicators of faecal contamination with other pathogens (e.g., rotaviruses, E. coli, Campylobacter) [43]. In conclusion, the waterborne outbreak described in the present study was most probably caused by faecal contamination of source A by groundwater, resulting from deficiencies in the maintenance of the water distribution system. Corrective measures were implemented based on the results of the investigation. To our knowledge, this is the first documented waterborne outbreak involving human C. coli infection. In-depth environmental investigations to identify the origin of water contamination are important in order to better understand microbial transmission from the environment and to implement appropriate controls and preventive measures. In addition, surveys on vehicles and vectors that transmit Campylobacter between hosts are crucial in order to better understand the epidemiology of these organisms.

Which contaminants or viruses or bacteria were found?

299

A large multi-pathogen waterborne community outbreak linked to faecal contamination of a groundwater system, France, 2000

A large waterborne outbreak of infection that occurred during August 2000 in a local community in France was investigated initially via a rapid survey of visits to local physicians. A retrospective cohort study was then conducted on a random cluster sample of residents. Of 709 residents interviewed, 202 (28.5%) were definite cases (at least three liquid stools ℠day or vomiting) and 62 (8.7%) were probable cases (less than three liquid stools ℠day or abdominal pain). Those who had drunk tap water had a threefold increased risk for illness (95% CI 2.4–4.0). The risk increased with the amount of water consumed (chi-square trend: p < 0.0001). Bacteriological analyses of stools were performed for 35 patients and virological analyses for 24 patients. Campylobacter coli, group A rotavirus and norovirus were detected in 31.5%, 71.0% and 21% of samples, respectively. An extensive environmental investigation concluded that a groundwater source to this community had probably been contaminated by agricultural run-off, and a failure in the chlorination system was identified. This is the first documented waterborne outbreak of infection involving human C. coli infections. A better understanding of the factors influencing campylobacter transmission between hosts is required. Keywords Campylobacter coli, epidemiology, gastroenteritis, norovirus, rotavirus, waterborne outbreak INTRODUCTION Contaminated drinking water causes extensive outbreaks of illness because of the large number of people served by water supply facilities. Detection of such outbreaks requires the identification of an increase in illness (usually the rate of gastrointestinal disease) in the exposed population and confirmation that water was the route of transmission [1]. Many countries have routine reporting systems for detecting foodborne and waterborne outbreaks of infection. However, in routine practice it is not easy to detect an increase in clinical cases and to link this increase to waterborne transmission; thus waterborne outbreaks are often unrecognised and underestimated [2]. Microbiological examination of water is more complex than examination of stools, and there is often a failure to detect pathogens in water. Norovirus, Cryptosporidium parvum, Giardia intestinalis and Campylobacter jejuni are the pathogens identified most frequently [3]. The present study describes a large waterborne community outbreak of infection with multiple pathogens, including Campylobacter coli. METHODS AND MATERIALS Background On 23 August 2000, a general practitioner (GP) informed the local district health department that 16 cases of acute gastroenteritis (AGE) had occurred among residents of a holiday camp in the Gourdon community since 21 August. Gourdon, which is located in the south-west of France, has a population of 4888 inhabitants, which increases by 50% in the July– August tourist season. Investigations at the holiday camp concluded that foodborne transmission was very unlikely, but found that tap water was highly contaminated with faecal coliforms. Preliminary information collected from physicians in Gourdon indicated an increase of the number of consultations for AGE. Gourdon tap water is supplied by two underground water facilities: facility A serving 3800 water consumers in Gourdon; and facility B serving 1200 water consumers in Gourdon and the surrounding communities. Water treatment in facility A consists of chlorination, which involves two automatic pumps working in rotation. The water then passes through two semi in-ground reservoir tanks (1100 m3 and 200 m3, respectively) before distribution. Facility B water is mixed with a fraction of facility A water in another 500 m3 semi in-ground reservoir tank before being distributed. This mixing with the treated water from facility A is intended to disinfect the water from facility B. Since the increase in AGE was noted only in Gourdon, contaminated tap water from facility A was suspected to be the origin of the outbreak. Therefore, the population of Gourdon was informed on 25 August to avoid the consumption of tap water or to boil it for 5 min. Survey of gastroenteritis-related medical consultations To confirm the outbreak, the 11 GPs in Gourdon and the local hospital emergency department were requested to provide the number of total daily visits and the number of cases of gastroenteritis cases occurring between 1 August and 11 September 2000. Retrospective cohort study A retrospective cohort study was performed on a sample of Gourdon households between 5 and 15 September 2000. Domestic households were selected randomly from the telephone directory. A definite case was defined as a Gourdon resident who had at least three liquid stools ℠day or vomiting between 1 August and 3 September, and who had been present in Gourdon for ‡ 1 day between 1 and 31 August; if less than three liquid stools ℠day or abdominal pain had occurred, the case was classified as probable. Exposure was defined as the consumption of tap water before 25 August 2000. A standardised telephone questionnaire was completed to obtain demographic data (gender, date of birth), food consumption (commercially prepared dishes, sea-food, raw milk, pastry), tap water consumption (before and after 25 August, daily number of glasses of water, consumption of boiled tap water), symptoms of gastroenteritis (date of onset, number of liquid stools ℠day, vomiting, nausea, abdominal pain, fever), medical care (in- or out-patient) and consequences on daily activities (sick leave, bed confinement) for each member of the selected household. Under the hypothesis that 60% of the population consumed tap water, an estimated 550 people were required for inclusion in the study to identify a relative risk (RR) of ‡ 3, with a 5% a risk and a power of 80%. Based on an average of two individuals per household (INSERM 1999 population census), 300 households were required for inclusion in the survey. The association between AGE and the consumption of tap water was assessed by calculating the risk ratio (RR) and 95% CI, taking into account a household cluster design effect. A dose-effect relationship was evaluated by analysing the RR trend for different categories of daily number of glasses of tap water consumed (chi-square trend). Proportions were compared using the corrected chi-square test, and means were compared using the Student test or Mann and Witney test. Data were analysed with Epi-Info 6.04c (CDC, Atlanta, GA, USA). The gender and age distribution of the cohort population were compared with that of the general population of Gourdon (INSERM 1999 population census) to assess whether it was representative. Under the hypothesis of representation, the total number of cases of gastroenteritis was estimated by applying the attack rate to the total population of Gourdon, including the tourist population (data provided by the Gourdon tourism office). The most likely period for contamination of the water distribution system was based on the epidemic curve and the minimal incubation period of the pathogens involved. Microbiological analyses of stools Stool samples were examined for the presence of Salmonella spp., Shigella spp., Staphylococcus aureus, Campylobacter spp., Yersinia enterocolitica, Escherichia coli, enterococci, rotavirus groups A and C, astrovirus, calicivirus, adenovirus types 40 and 41, enterovirus, hepatitis A virus and Cryptosporidum. Viruses were detected using an enzyme immunoassay for group A rotavirus, astrovirus and adenovirus 40 ℠41, and by RT-PCR for calicivirus, group C rotavirus, enterovirus and hepatitis A virus [4–6]. Campylobacter isolates were sent to the National Reference Centre for Campylobacter for identification and fingerprinting. Species identification was by means of phenotypic tests (API Campy; bioMe´rieux, Lyon, France), with use of PCR to differentiate Campylobacter jejuni from Camplylobacter coli [7]. Random amplified polymorphic DNA fingerprinting was performed with primer 3881, 5¢-AACGCGCAAC. Environmental investigation An environmental and microbiological investigation of the water distribution system was performed. The inspection concerned the water catchments, the water distribution and sewage system, agricultural practices and private households around the sources. Bacteriological (Escherichia coli, enterococci, faecal streptococci, thermotolerant coliforms, sulphite-reducing clostridia) and chemical (chlorine) analyses were performed on tap water and on the water sampled from the two groundwater sources. After 23 August, the water chlorine concentration and the presence of coliform bacteria were monitored daily at several sites of the water system before and after the resort treatment. Examinations for viruses (group A rotavirus, astrovirus, norovirus, enterovirus, hepatitis A virus) and Campylobacter spp. were undertaken on two 10-mL tap water samples collected inhouse on 23 August, and on two 10-mL samples from groundwater sources A and B on 25 August. Viruses were detected by RT-PCR, followed by hybridisation with a specific primer to improve sensitivity, while C. jejuni and C. coli were detected by PCR with specific primers. Group A rotavirus isolates detected in stools and in water were sequenced using Beg 9 and End 9 primers [8] and compared using the Infobiogene GCG software (http://www.infobiogen.fr). Bacteriological (revived 36C, revived 22C, thermotolerant coliforms, streptococci and faecal enterococci), physical (temperature, conductivity) and chemical (oxygen consumption, ammonium, nitrates, phosphorus) parameters from source A and the river were analysed. Trends were compared for an 8-week period between 21 December 2000 and 7 February 2001. RESULTS Epidemiology Between 1 August and 11 September 2000, 7104 Gourdon residents consulted a GP or the local hospital emergency department. Between 24 and 28 August, gastroenteritis accounted for 44% (479 ℠1097) of all medical visits, compared with 6% (126 ℠2185) during the first 15 days of August (Fig. 1). Among the 498 households contacted for the household survey, 198 were excluded (103 did not respond, 49 did not fulfil the inclusion criteria, 43 refused to participate, and three for unknown reasons). The questionnaire was completed for 300 households (709 individuals). Data were missing concerning gender for one individual and age for 11 individuals. Overall, 331 (46.7%; 95% CI 44.5–49.0) of respondents were male and 377 (53.2%; 95% CI 50.9–55.5) were female. The median age was 44 years (range < 1–94 years). The age and gender distribution of the sample was similar to that of the general population. During the outbreak period, 264 of the 709 respondents (attack rate (AR) 37.2%) had been ill, with 202 definite cases (AR 28.5%; 95% CI 24.6– 32.4) and 62 probable cases (AR 8.7%; 95% CI 6.5– 11.0). The AR was higher in females (31.8%) than in males (24.8%), and was highest in children aged < 6 years (42.1%, Table 1). The epidemic curve (Fig. 2) suggests that the outbreak started between 14 and 19 August. The shape of the epidemic curve in the retrospective cohort study was very similar to that based on the number of visits to GPs for gastroenteritis. Patients complained mainly of diarrhoea, abdominal pain and nausea (Table 2). No cases of bloody diarrhoea were notified. The mean duration of symptoms was 4.5 days (median 3, range 1–30 days). Among definite cases, 52% had visited a GP, 42% had taken sick leave for ‡ 1 day, 32% had required bed rest, and 2.9% (six patients) had been hospitalised. In total, 336 (47.4%) respondents had drunk tap water before 25 August, whereas 646 (91.1%) did not drink tap water after 25 August. The risk for illness, taking into account the sample cluster design effect of 1.4, was three-fold greater (95% CI 2.2–4.1) for those who had drunk tap water before 25 August, and increased with the amount of water consumed (chi-square trend: p < 0.0001; Table 3). The risk of gastroenteritis associated with drinking tap water did not change substantially after adjustment for age (RRMH 3.3, 95% CI 2.5–4.4), and there was no modification effect according to age. Food consumption was not associated with illness. Applying the AR of 37% to the total population of Gourdon exposed to the water supply (4888 residents + 2200 tourists), the number of individuals affected by this outbreak was estimated to be 2600 (95% CI 2400–2900), assuming an exposure to tap water and a risk for gastroenteritis among the tourists similar to that among the inhabitants. Microbiological stool analyses Among the 35 stool samples analysed for bacteria, 11 (31.5%) were positive for C. coli. Twenty-four of the 35 samples were also tested for viruses: rotavirus A was detected in 17 (71.0%) and norovirus in five (21.0%) samples, while nine (37.5%) samples were co-infected (four C. coli + rotavirus; one C. coli + norovirus; three rotavirus + norovirus; and one C. coli + rotavirus + norovirus). Of the 11 C. coli isolates, six were characterised as biotype 2, and were identical following RAPD analysis, while the remaining five were of biotype 1. Rotavirus typing by RTPCR identified a single genotype P-(8), G1. Molecular characterisation of the norovirus isolates identified two different patterns, with three isolates being identical and close to the Saragota strain (genogroup I) and two close to the new GGIIb variant (genogroup II). Environmental investigation All of the water samples from source B were negative, regardless of the site of the distribution system tested. Analyses of the water originating from source A, sampled after leaving the water-works on 23 and 25 August, revealed high concentrations of E. coli (2263 / 100 mL), enterococci (415 / 100 mL), total coliforms (90 / 100 mL), faecal streptococci (136 / 100 mL) and sulphite-reducing clostridia (10 / 100 mL), and an absence of chlorination. Rotavirus A genotype G1 was detected in the crude water sampled from source A. The analyses for enterovirus, astrovirus, norovirus, hepatitis A virus and Campylobacter were negative. Source A was situated in a valley containing a river, below a hamlet of 15 dwellings, including farms breeding cattle, pork and sheep. The source was not supposed to be supplied by the river. The river resulted from the convergence of about ten different streams, with one originating close to a slaughterhouse, and was exposed to agricultural run-off and manure effluent; sheep and cattle grazed in meadows bordering the river, and a large amount of sewage from water treatment facilities was sprayed on to land close to the river. Cowsheds were close to the water catchment area. The immediate protected area around source A was too small and littered with building rubbish. Air vents without protection in the door made the source accessible to animals and insects. Bacteriological analyses of the water samples from source A and the river revealed that the contamination was of both human and animal origin, with a higher concentration in the river. The levels and trends of the chemical parameters of source A were similar to those of the river, strongly suggesting that the river was connected to source A (Fig. 3). A retrospective review of routine data collected between 1991 and 2000 showed irregularities in the maintenance of the water distribution system and the water treatment plant, particularly during the month of August. In the hamlet, the sewage system of the 15 households did not comply with sanitary rules. All but three ejected waste either into a cesspool or superficially without treatment. Three regions of gene 9 of the rotavirus detected in stools and water were compared. Sequencing showed that the strains were closely related but not identical. Assuming a likely onset of the outbreak between 14 and 19 August, and taking into account a minimum incubation period of 2 days for rotavirus A, norovirus and C. coli, the contamination of the water distribution system probably occurred between 12 and 17 August. Contamination probably persisted until 28 August, as intensive chlorination, beginning on 25 August, was effective only from 28 August onwards (Fig. 2). DISCUSSION The present descriptive and retrospective cohort studies confirmed that the outbreak of infection in the community of Gourdon was related to the consumption of tap water and affected 2600 individuals. The outbreak involved multiple pathogens, with group A rotavirus and C. coli being the most frequent, followed by norovirus. Results of environmental investigations suggested that water contamination was related to the contamination of the source, combined with a failure of the water chlorination system. The association with tap water consumption, the dose-effect relationship, and the fact that multiple pathogens were involved, were all strongly in favour of tap water as the origin of infection. Moreover, the number of cases decreased dramatically after the water restriction advice and the implementation of control measures on the water supply. No food products explored in the study were associated with illness. An AR of 14.5% and 31% was observed among adults and children aged < 6 years, respectively, who reported that they had not consumed tap water. This relatively high rate could be explained by several phenomena. Many individuals may have consumed tap water without being aware of it, e.g., in drinks or ice cubes consumed outside of the home, via raw fruits and vegetables, dishes prepared with unboiled water, or while cleaning their teeth. It can be hypothesised that people who became ill without being aware of having consumed tap water were exposed to small amounts of water and had been subjected to a small infective dose. In addition, children, more than adults, may have been contaminated through person-to-person transmission (e.g., family contact), which is the most common route of transmission for rotavirus and calicivirus. Waterborne contamination can affect large numbers of individuals [3,9–11]. In the present study, assuming a 50% increase in the population during the tourist season, the number of affected people was estimated to be between 2400 and 2900. The cohort study showed that almost half (46.5%) of the cases consulted a GP. Applying this proportion to the estimated number of cases yielded an estimated 1116–1349 ill individuals who consulted a GP for gastroenteritis, similar to the number of consultations for gastroenteritis actually reported by the GPs (n = 1037). The stool sample analyses revealed the presence of multiple pathogens (rotavirus A, norovirus genogroups I and II and C. coli) as single or co-infections. Except for C. coli, these pathogens have been implicated in numerous waterborne outbreaks, suggesting an influx of sewage into the water supply system [10,12–18]. Rotaviruses are known to be responsible for large epidemics during winter seasons, affecting particularly children aged < 3 years. These epidemics are caused by the faecal route of contamination (personto-person transmission). The high proportion of stools positive for rotavirus (71%) is uncommon in waterborne outbreaks and probably reflects massive contamination [15,16,19]. Different molecular sequences of rotavirus A genotype G1 found in stools and water indicated a human faecal source of infection. In the present study, contamination of water by sewage could be linked with the large amount of sewage from the water purification plant sprayed on to land close to the river and the water catchments. In this outbreak, C. coli, which was identified in 32% of stools, was the second most frequent pathogen. To our knowledge, human C. coli contamination has not been described previously as a causative agent in waterborne outbreaks, unlike C. jejuni [11,20–25], although C. coli has been isolated from water [26]. The main symptoms observed (diarrhoea, abdominal pain, nausea) and the mean duration of the disease (4.5 days) were compatible with a Campylobacter infection. No bloody diarrhoea was reported, although bloody diarrhoea is not constant in Campylobacter infections [27]. The high proportion of individuals who had to take sick leave, and the fact that six individuals were hospitalised, suggested a relative severity of symptoms, which is much more compatible with Campylobacter infection than with virus infections. Among the hospitalised patients, 60% were infected with C. coli. Among patients whose stools were analysed and whose age was known, cases infected only with rotavirus were younger (mean age 49 years; range, 5 months to 95 years) than cases infected only with Campylobacter (mean age 66 years; range 23–88 years) or co-infected with rotavirus plus Campylobacter (mean age 69 years; range 2–91 years). Moreover, the proportion (52%) of medical visits was unusually high compared with the proportion of visits observed during other waterborne outbreaks [13,16,28]. Campylobacter needs specific atmospheric conditions for growth and survival and its detection can be very difficult [29,30]. Because of a rapid physiological transformation to a viable, though non-cultivable, state, it has been suggested that conventional culture methods often fail [31]. Nevertheless, in the present study, the failure to detect C. coli in water, and the detection of noroviruses in stools, could be caused by the small (10 mL) water samples examined. Ha¨nninen et al. [30] proposed that the volume used for detection of suspected pathogens in drinking water should be at least 8–10 L. Groundwater is considered widely to be microbiologically clean, and is usually used for drinking without treatment. However, groundwater can be contaminated by the environment [26,32]. Stanley et al. [33] showed that Campylobacter can occur in groundwater, with hydrological evidence suggesting that the source of contamination in their study was a dairy farm situated within the hydrological catchment area of the groundwater, which was confirmed when identical strains of C. jejuni were isolated from groundwater and the dairy herd. In the present study, Campylobacter was not identified from groundwater, and dairy herds and other domestic animals in farms around the source were not investigated. However, results of bacterial, chemical and physical water analyses suggested strongly that a river contaminated by agricultural run-off from surrounding sheep and cattle grazing meadows was contaminating the source. In addition, the immediate area of the source was insufficiently protected and vulnerable to the environment. Wild birds, rodents, pigs and cows are all hosts for C. coli and could contaminate surface water easily [26,27,34,35]. Campylobacters are sensitive to many environmental factors and to chlorine [33,36,37], but a failure of the chlorination system could have resulted in the presence of a high concentration of C. coli. Waterborne outbreaks are often caused by several factors that act together [18,20,23,38]. In some outbreaks, heavy rainfall has been responsible for increasing agricultural run-off into rivers, although this did not occur before the outbreak in Gourdon. This underlies the need for increased frequency of routine water quality monitoring, chlorination of the water supply according to environmental conditions, and implementation of preventive measures. This outbreak raised the question of the delay in raising the alert. An outbreak can be recognised by: (i) the poor quality of water detected during regular controls; (ii) complaints about water quality; (iii) an incident in the supply and ℠or sewage water system during maintenance work; and (iv) an increased number of cases of gastroenteritis. In most instances, waterborne outbreaks are suspected when there are already a large number of declared cases [3,12,18,28,38,39]. In the present outbreak, the initial warning was given by a GP who notified officials 5–10 days after the assumed date of onset. This highlights the role of GPs in the detection of an unusual event, particularly during the summer, when an increase in acute gastroenteritis is less expected than during winter [40,41]. In addition, a routine water quality monitoring system able to detect contamination of drinking water before the occurrence of illness and complaints about the water quality should be used as an early warning source. The retrospective cohort study of Gourdon residents allowed the scale and severity of the outbreak to be quantified precisely and the hypothesis that tap water was the vehicle of the outbreak to be tested. When provided with a cohort sample representative of the study population, the number of people affected can be estimated. However, as the exposure of the population was high, a relatively large sample size was needed to test the association between the disease and the exposure. Because of misclassification of exposure (unrecognised consumption of tap water), the attributable risk is probably underestimated. In the absence of obvious alternative explanations, water contamination is the most likely common cause of an explosive outbreak of AGE occurring in a large population. Tillett et al. [42] proposed categorising levels of evidence for waterborne outbreaks. A descriptive study suggesting a water-related outbreak, together with the identification of the same pathogen in stools and water, should be sufficient to conclude that an outbreak is associated strongly with water consumption. Faecal indicators of water quality are currently insufficiently sensitive for some pathogens, notably viruses and parasites (e.g., Cryptosporidium). Therefore, in the absence of evidence from an analytical study to demonstrate an association, the link between illness and water can only be classified as probable or possible. However, total coliform counts are good indicators of faecal contamination with other pathogens (e.g., rotaviruses, E. coli, Campylobacter) [43]. In conclusion, the waterborne outbreak described in the present study was most probably caused by faecal contamination of source A by groundwater, resulting from deficiencies in the maintenance of the water distribution system. Corrective measures were implemented based on the results of the investigation. To our knowledge, this is the first documented waterborne outbreak involving human C. coli infection. In-depth environmental investigations to identify the origin of water contamination are important in order to better understand microbial transmission from the environment and to implement appropriate controls and preventive measures. In addition, surveys on vehicles and vectors that transmit Campylobacter between hosts are crucial in order to better understand the epidemiology of these organisms.

Which were the symptoms?