Datasets:

Galahad3x

/

QADatasetForPatho

like 0

0

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

What happened?

1

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

What was the event?

2

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

When did this happen?

3

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

When did this event start?

4

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

What is the date of this event?

5

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

How long was the event?

6

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

How long did the event last?

7

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

In which street did this happen?

8

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

In which city did this happen?

9

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

In which region did this happen?

10

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

In which country did this happen?

11

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

In which country did this happen?

12

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

Where did this happen?

13

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

What caused the event?

14

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

What caused the dried custard mix may have been inadvertently reconstituted with contaminated water?

15

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

What was the cause of the event?

16

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

What source started the event?

17

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

How was the event first detected?

18

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

How many people were ill?

19

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

How many people were hospitalized?

20

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

How many people were dead?

21

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

Which contaminants or viruses or bacteria were found?

22

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

Which were the symptoms?

23

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

What did the patients have?

24

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

What were the first steps?

25

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

What did they do to control the problem?

26

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

What did the local authorities advise?

27

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

What were the control measures?

28

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

What type of samples were examined?

29

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

What did they test for in the samples?

30

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

What is the concentration of the pathogens?

31

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

What steps were taken to restore the problem?

32

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

What was done to fix the problem?

33

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

What could have been done to prevent the event?

34

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

How to prevent this?

35

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

What were the investigation steps?

36

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

What did the investigation find?

37

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

How was the infrastructure affected?

38

A community outbreak of food-borne small round-structured virus gastroenteritis caused by a contaminated water supply

SUMMARY In August 1994, 30 of 135 (23%) bakery plant employees and over 100 people from South Wales and Bristol in the United Kingdom, were affected by an outbreak of gastroenteritis. Epidemiological studies of employees and three community clusters found illness in employees to be associated with drinking cold water at the bakery (relative risk 3±3, 95%, CI 1±6–7±0), and in community cases with eating custard slices (relative risk 19±8, 95%, CI 2±9–135±1) from a variety of stores supplied by one particular bakery. Small round-structured viruses (SRSV) were identified in stool specimens from 4 employees and 7 community cases. Analysis of the polymerase and capsid regions of the SRSV genome by reverse transcription-polymerase chain reaction (RT-PCR) demonstrated viruses of both genogroups (1 and 2) each with several different nucleotide sequences. The heterogeneity of the viruses identified in the outbreak suggests that dried custard mix may have been inadvertently reconstituted with contaminated water. The incident shows how secondary food contamination can cause wide-scale community gastroenteritis outbreaks, and demonstrates the ability of molecular techniques to support classical epidemiological methods in outbreak investigations. INTRODUCTION The small round-structured virus (SRSV or Norwalklike virus) is the most common cause of epidemic food- and water-borne viral gastroenteritis [1]. Several large outbreaks due to contaminated drinking water have been described [2–4]. Food-borne outbreaks fall into two main categories according to the source and mode of contamination. They arise either when food contaminated with virus arrives in the kitchen, or when food is contaminated in the kitchen by an infected food handler [5]. Primary contamination is almost invariably associated with molluscan shellfish such as oysters and can cause wide-scale outbreaks [6–8]. By contrast, a range of foods may be secondarily contaminated by food handlers, but outbreaks tend to be well circumscribed, and community outbreaks are uncommon [9]. We describe one of the largest food-borne community outbreaks of this kind identified in the United Kingdom. THE OUTBREAK On 10 August 1994, the personnel department of a large bakery in South Wales notified the local Environmental Health Department (EHD) that 6 of its 135 staff were ill with vomiting and diarrhoea. The premises were inspected and affected employees advised to stay off work for 48 h after symptoms had resolved. On 15 August, the EHDwas informed by two neighbouring EHD of 16 community cases of gastroenteritis all of whom had eaten cakes from outlets of the bakery’s retail chain. We undertook an investigation in order to describe the outbreak, identify cases, examine risk factors for illness and determine the cause. METHODS Epidemiological investigation A retrospective cohort study of all 135 bakery employees who had been at work between 1 and 18 August 1994 was conducted. Structured questionnaires were used to obtain details of any illness, shiftwork pattern, work responsibilities, canteen use and foods consumed. A case was defined as any employee with either diarrhoea (3 or more loose stools in 24 h) or vomiting during the above period. Case ascertainment to identify community cases was carried out by alerting all EHDs in South Wales via EPINET (an electronic mail network used for epidemiological surveillance), local general practitioners by telephone, and the general public by means of a news release. Stool specimens recently submitted to the local laboratory were also reviewed and a survey of all bakery outlets in the retail chain undertaken to identify any customers or staff with recent gastrointestinal illness. A case was defined as an individual with either diarrhoea or vomiting and a date of onset between 12 and 14 August, and a cluster as two or more related cases. Cohort studies were carried out on three clusters of community cases. These had occurred among groups of office or factory workers who had purchased a cake selection from different outlets of the bakery’s retail chain. Subjects were interviewed using a structured questionnaire to ascertain details of illness and food histories. We calculated risks with 95% confidence intervals (CI) to measure association of illness with individual risk factors and used the χ# test with Yates’ correction or Fisher’s exact test to derive P values. Multiple logistic regression was performed using ï§ï¬ï©ï­ statistical software [10] to fit a sequence of models to assess the relationships between illness and risk factors, taken separately and together, to allow for confounding. Environmental investigation The bakery plant was thoroughly inspected, particularly the food production areas, water system, toilet and washing facilities, and the staff canteen. Details of food production were obtained and food hygiene policies and procedures reviewed. In-depth interviews of eight key workers were undertaken. Laboratory investigation Stool samples were available for 38 ill bakery employees and 27 community cases. Samples were cultured for bacteria (including Salmonella, Shigella and Campylobacter species) and faecal smears examined by microscopy for oocysts or parasites of Giardia and Cryptosporidium species. Stool samples from 7 bakery employees and 17 community cases (obtained within 48 h of onset of symptoms) were prepared as 10–20% emulsions in Eagle’s minimal essential medium and examined for viruses by solidphase immune electron microscopy (SPIEM). SRSV were captured onto formvar}carbon grids coated with anti-human IgG and human convalescent serum by placing the grid onto a small drop of faecal emulsion for 2 h at room temperature. The grid was then washed with water (5 s), stained using 1±5% phosphotungstic acid (3 s), dried and examined using a Philips EM420 electron microscope (¬65 000). Samples were also analysed by reverse transcription-polymerase chain reaction (RT-PCR). Viral RNA was extracted from faecal emulsions by RNA enrichment [11] followed by concentration and purification [12]. One hundred µl of emulsion was extracted in 500 µl of Tri-reagent (Sigma-Aldrich Company Ltd, UK) by vortex mixing for 10 min, followed by addition of 100 µl of chloroform and vortexing for a further 10 min and finally, centrifugation (13 000 g) for 15 min. The upper phase was transferred to a fresh tube containing 1 ml L6 buffer [12] and 40 µl of silica slurry and the sample processed as described previously [12]. The purified RNA was eluted with 50 µl of sterile distilled water. For cDNA synthesis, 36±5 µl of the RNA solution was made up to a 50 µl reaction containing 50 mï­ Tris-HCl (pH 8±3), 75 mï­ KCl, 10 mï­ DTT, 3 mï­ MgCl# , 0±5 mï­ dNTPs, 1 pg pdN' primers (Pharmacia) and 100 units MMLV reverse transcriptase (Gibco-BRL). The reaction was incubated at 37 °C for 1 h. Two µl of cDNA was used to seed each 20 µl PCR reaction containing 10 mï­ Tris-HCl (pH 9±0), 3 mï­ MgCl# , 50 mï­ KCl, 0±1% Triton X100, 0±2 mï­ dNTPs and 0±1 µg of either Group 1 or Group 2 SRSV primers [13]. PCR cycling conditions were 35 cycles at 94 °C for 20 s (3 min cycle 1), 50 °C for 20 s, 72 °C for 5 s on a Perkin–Elmer 9600. For the capsid primers, the extension phase at 72 °C was increased to 10 s. Fifteen ll of PCR product was mixed with 2 ll of loading dye (50% glycerol in TBE running buffer containing 0–05% bromophenol blue) and loaded onto a TBE-2% agarose gel. SRSV positive specimens were identified by the presence of a 155 bp amplicon for the polymerase primers or a 223 bp amplicon for the capsid primers. For DNA sequencing, 2 ll of cDNA from PCR positive specimens were amplified in 100 ll preparative reactions by simply scaling up the above PCR reactions. Amplicons were purified by PEG precipitation [14] to remove primer dimers and unincorporated nucleotides, resuspended in water, and chloroform extracted to remove residual traces of PEG. DNA sequencing was carried out using the Dye Deoxy method on an ABI377 automated DNA sequencer (Applied Biosystems). DNA sequence from both strands was obtained by using the same primers as used to produce the amplicon. For polymerase gene amplicons, the DNA product was initially purified by excising the band from TBE agarose gel using a sterile disposable scalpel. The excised band was placed into a screw top microfuge tube containing 1 ml of sterile water and incubated at 100 °C for 10 min to melt the agarose. Preparative PCR was carried out as above using 2 ll of the gel purified product to seed the reaction. Consensus DNA sequences were determined by comparing the sequence of both DNA strands and ambiguities resolved by manual inspection of the electropherograms. Sequences were aligned and phylogenetic trees generated using the programmes clustal w [15], retree [16] and treeview [17]. Five sets of water samples and environmental surface swabs were taken at the bakery for microbiological examination during August–October. In addition, microbiological results were available from routine water and environmental samples collected by the bakery on 10 and 12 August and tested at a private laboratory. Virological investigations were not carried out on these samples. RESULTS Epidemiological investigation Questionnaires were completed by 133}135 bakery employees of whom 30 (23%) fulfilled the case definition. The first case occurred on 1 August, with an epidemic peak between 9 and 12 August (Fig. 1). Symptoms included diarrhoea (80%), abdominal pain (80%), vomiting (70%), fever (43%) and headache (43%). Median duration of illness was 48 h (range 14–168 h) and no one required hospitalization. Attack rates were similar in both sexes (25% male vs. 17% female) and both work shifts (22% day shift vs. 25% night shift). There was no relationship between risk of illness and work undertaken in different locations within the bakery, nor with eating bakery products. However, employees were more likely to be ill if they drank cold water either from the drinking fountain (relative risk 3–0; 95% CI 1–5–5–9) of from any other source in the bakery (relative risk 3–3; 95% CI 1–6–7–0) during the week of the outbreak (Table 1). Employees from both day and night shifts ate breakfast in the bakery canteen, and eating breakfast on 8 August (relative risk 19–2; 95% CI 2–5–148–0) or 9 August (relative risk 17–2; 95% CI 2–2–137–9) was associated with becoming ill within 16–48 h of the meal. No illness was reported among canteen staff themselves in the week leading up to 10 August. Illness was associated with eating any one of a wide range of cooked foods for breakfast on 8 or 9 August including sausages, fried eggs, bacon, toast, beans and tomatoes, as well as with drinking tea or coffee (Table 1). Using multiple logistic regression to control for drinking water and for the range of foods eaten, only consumption of tomatoes (adjusted odds ratio 6–4; 95% CI 1–0–40–4) and water from the drinking fountain (adjusted odds ratio 4–0; 95% CI 1–7–9–6) on 8 August were associated with illness. Regression models also suggested a significant association between illness and drinking tea or coffee on 9 August. However, only a sub-set of cases were eligible or had148 R. Brugha and others Table 1. Activity-specific attack rates for bakery employee cohort (nfl133) sufficient data for these analyses and the models were unstable, despite the removal of non-significant variables. Consequently, they could not be reliably interpreted. One hundred and four community cases from six districts across South Wales and the West Country were identified. The largest single cluster was nine cases. Median age of cases was 40 years (range 11–77 years) and 58% were women. All but four of the cases had bought custard slices on 11 August from 17 of the bakery’s 42 retail outlets and all became ill between 12 and 14 August. Median incubation period was 32 h (range 10–71 h). All 42 persons in the three community groups were interviewed and 19 met the case definition. The main symptoms were diarrhoea (89%), abdominal pain (79%), vomiting (74%), headache (63%), and fever (58%). Eating a custard slice produced by the bakery was significantly associated with illness (relative risk 19–8; 95% CI 2–9–135–1) (Table 2). Eighteen of 20 persons who reported eating a custard slice developed gastroenteritis, a 90% attack rate. Environmental investigation The bakery plant is of modern design (Fig. 2) and responsible for supplying 42 retail bakeries. Custard slices were made daily in the cream room using a cold custard mix that required no cooking. The custard mix was prepared by day staff from reconstituted custard powder (using water from the public water supply drawn in the cream room) which was mechanically mixed, scooped onto pastry sheets with a metal scoop and levelled with a palette knife before applying a second pastry layer. The slices were kept chilled and finished by night staff who applied pre-prepared fondant icing and cut the slices in the confectionery room. The finished product was placed in trays, taken to the chill despatch room and loaded onto delivery vans. Between 2600 and 3000 custard slices were distributed daily. In-depth interviews of bakery employees established that Worker 1, shortly after arrival at work on 10 August (the date the suspect batch was produced), developed sudden, explosive diarrhoea and vomiting. He vomited in the toilet bowl, flushed the toilet, washed his hands, reported the incident to his manager and was sent home on sick leave. Worker 2, who was involved in preparation of custard slices on the same day, admitted applying the cold custard mix to the pastry by hand rather than using the designated palette knife. He became ill at home on the following evening, 11 August, at 8.00 p.m. and went on sick leave. Both workers had eaten breakfast in the staff canteen on 8 and 9 August. Staff interviews also revealed a number of unhygienic practices at the bakery. These included unprotected food handling, use of a fire exit between the staff canteen and bakery production area at night (by-passing the hand washing area) (Fig. 2), poor temperature control of water for hand washing and malfunctioning soap dispensers in the toilets. Laboratory investigation One community case was positive for Salmonella mbandaka and samples from two bakery employees yielded Campylobacter sp. and enterotoxogenic Staphylococcus aureus, respectively. SRSVs were identified by SPIEM in samples from 1 bakery employee and 3 community cases. However, SRSV RNA was detected by RT-PCR in 6 employees and 8 community cases using capsid primers, and in 7 employees and 12 community cases using polymerase primers. RT-PCR is more sensitive than EM in detecting SRSV and, in our experience, the polymerase primers have a better detection rate than the capsid primers. However, the polymerase primers are less discriminatory, and consequently, genotype information is not always resolved (Table 3). Nucleotide sequences from the polymerase region were obtained for 4 employees and 7 community cases (Fig. 3). Two employees had identical group 1 virus sequences and 2 other employees had extremely similar group 2 sequences (99% identity). Three community cases had group 1-like sequences which were distinct from each other and from the sequences identified in bakery employees. The sequence of the polymerase amplicon from the virus in the stool of one consumer (Bristol A) proved particularly difficult to resolve. It possessed similarities to sequences identified in both employees and community cases and may indicate a mixed infection of genogroup 1 viruses that were co-amplified by the polymerase primers. One group 2 sequence from a community case (Swansea F) was identical to that identified in an employee (Baker F) and had 99% identity with a sequence from Baker G. Three other community cases with a different genogroup 2 virus had identical (Bristol D and Cardiff E) or similar (Swansea A; 99% identity) nucleotide sequences. The possibility that community cases carried mixed infection was investigated by sequencing capsid amplicons (Fig. 4). This confirmed that three community cases were carrying both group 1 and group 2 viruses. Two cases (Cardiff C and Swansea H) had group 1 viruses with similar sequences (96–8% identity), but the third case (Cardiff E) carried a completely different sequence type (69–5% identity). All three cases had group 2 viruses which were closely similar, two being identical (Cardiff E and Swansea H) and the third (Cardiff C) showing 98–8% identity. A further two amplicons were sequenced for group 1 virus from an employee (Baker A) and a community base (Bristol A) and were found to be identical to each other, but distinct from the sequence types carried by the community cases with mixed infections (Fig. 4). A sample of water obtained from the drinking fountain in the food production area on 10 August grew coliforms"1200}100 ml, Escherichia coli" 1200}100 ml and faecal streptococci 160}100 ml. Unfortunately, this water sample was not available for virological analysis. Environmental swabs from numerous food preparation areas and sinks also taken on 10 August were all negative. Water samples taken on 12 August from the same drinking fountain, the mains supply, a chilled outlet and the cream room supply (where the custard slices were produced) grew no coliforms or faecal pathogens. Further samples taken between August and October from the mains water supply and from various water outlets and food preparation areas at the bakery were all negative. DISCUSSION This is one of the largest community outbreaks of food-borne viral gastroenteritis due to secondary food contamination to be reported in the United Kingdom. It followed an apparent point source outbreak of gastroenteritis at a large bakery plant that peaked on 10 August. The epidemic curve, clinical features, and detection of SRSV in stool confirm this outbreak as viral gastroenteritis. Epidemiological studies identified custard slices as the vehicle of transmission, and detection of SRSV of identical nucleotide sequence in samples from bakery employees and community cases imply a common source outbreak. The identification of multiple SRSV strains in both bakery employees and community cases, and of mixed infections in community cases, strongly supports the view that the source of the outbreak was contamination of the bakery water supply. Mixed infections have only previously been reported in outbreaks associated with the consumption of shellfish [18] or from bathing in sewage contaminated recreational water [19]. The hypothesis that the bakery outbreak was most probably caused by contaminated water supply is supported by epidemiological evidence of an association between illness among bakery employees and drinking cold water at the bakery; microbiological evidence of heavy faecal contamination of the drinking fountain on 10 August; and by illness due to multiple SRSV strains. Alternatively, the outbreak might have been caused by a vomiting incident in the bakery canteen with secondary aerosol spread, but no such event was reported. Nor can the outbreak be explained by use of a contaminated toilet since both sexes were similarly affected in spite of the existence of separate male and female toilets. Finally, neither of these explanations can explain illness due to multiple SRSV strains. The community outbreak was associated with eating custard slices. These were most probably contaminated by the use of contaminated water to reconstitute the cold custard mix. This hypothesis is supported by the presence of mixed infection due to both SRSV genogroups and several genotypes in community cases. Classically, food-borne spread of SRSV follows direct contamination by a food handler. This is what occurred in one previously reported bakery-associated outbreak where an ill employee was found to have contaminated cakes by submerging his arms in buttercream frosting [20]. Although the employee who prepared the custard slices reported illness, his date of onset was 11 August, the day after the implicated batch was produced. The employee may have incorrectly reported his illness onset date, but we found no evidence to contradict his account. Alternatively, the custard slices might have been contaminated by pre-symptomatic excretion of virus [21, 22] or by direct virus transfer [23], as have been suggested in previous outbreaks. However, none of these explanations account for the virological findings. Previous molecular studies of SRSV outbreaks have been able to show identical nucleotide sequences in all cases in point source food-borne outbreaks [24] or in index and secondary cases in outbreaks involving spread from person-to-person [25, 26]. Sequence heterogeneity among SRSV is such that identical sequences in outbreaks strongly support a common source of infection. Co-infection with both genogroup 1 and 2 has occasionally been reported [18, 19], but only as a consequence of consumption of contaminated shellfish (which are filter feeders able to concentrate a mixture of viruses in their mid gut gland) [18] or from bathing in contaminated recreational water [19]. In this outbreak, molecular biology identified both SRSV genogroups, with several genotypes in each group, among both bakery employees and community cases. Some of the sequences identified in bakery employees were identical to some of those found in community cases but there were some sequence types unique to bakery employees and some unique to community cases (although not every community case was studied at the molecular level). The absence of mixed infections in the few bakery employees that were tested by PCR may either reflect prior immunity to some virus strains or may be due to infection by secondary spread from workmates (3 of 4 employees in whom genogroup identification was successful all had late illness onset dates). No faecal sample was available from the bakery employee who could have contaminated the custard slices. By contrast, three community cases had mixed genogroup infections. Altogether, six different genotypes (3 group 1 and 3 group 2) were identified among community cases from the polymerase region (assuming even small variations in sequence are significant), and six genotypes (4 group 1 and 2 group 2) from the capsid region. This suggests that the custard slices may have been contaminated by at least seven SRSV sequence variants. Some of the sequence variations are small (around 1%) and may be a consequence of Taq polymerase-induced differences. However, even if only gross variations are considered as significant there remains evidence of infection of community cases by at least 4 or 5 distinct sequence types. Additionally, the community cases with mixed infections had both genogroup 1 and 2 in their stools. How did the water supply become contaminated? At the time of the outbreak, no obvious explanation for possible contamination of the bakery water supply was found. Subsequently, during investigation of an outbreak of gastroenteritis among employees of another premises on the same industrial estate, an unauthorized connection between river water and the mains supply without isolating valve control was identified at a nearby factory. It is possible that this connection may have allowed intermittent reverse flow contamination of mains water supply when mains water pressure was unusually low. A similar situation has been described due to reverse flow between a farm’s water irrigation system and the drinking water supply that caused gastrointestinal illness in a local rural community [27]. Unfortunately, none of the contemporary bakery water samples were available for virological analysis. The batch of custard slices that caused the community outbreak was prepared by the night shift on 10 August and distributed to 42 retail bakeries early the following morning. Over 2600 custard slices were sold from the contaminated batch, so as many as 2300 cases may have occurred (assuming a uniform 90% attack rate). However, only 104 cases were reported and only 17 of 42 retail outlets implicated. A variable attack rate due to variable contamination of custard slices is unlikely if the source was contaminated water. Extensive undernotification of cases may therefore have occurred. This outbreak demonstrates the value of molecularFood-borne community gastroenteritis 153 techniques in the investigation of SRSV outbreaks. It also highlights the potential for wide-scale outbreaks of food-borne viral gastroenteritis even when the products and premises involved are generally regarded as low risk. As the scale of production and distribution of food products grows, the potential for large outbreaks which cross both district and national boundaries also increases. Several international foodborne outbreaks have been reported in recent years [28]. This outbreak illustrates not only the hazards of secondary viral contamination of food but also serves to emphasize the importance of good cross-boundary communication to ensure effective outbreak management and control. ACKNOWLEDGEMENTS We are indebted to the staff of Taff-Ely Borough Council Environmental Health Department and of many other Environmental Health Departments throughout South Wales for assistance with the investigation, to Cardiff and Bristol Public Health Laboratories for undertaking microbiological analysis and to Dr A. V. Swan of the PHLS Statistics Unit for statistical advice and assistance.

What were the infrastructure complaints?

39

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

What happened?

40

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

What was the event?

41

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

When did this happen?

42

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

When did this event start?

43

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

What is the date of this event?

44

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

How long was the event?

45

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

How long did the event last?

46

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

In which street did this happen?

47

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

In which city did this happen?

48

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

In which city did this happen?

49

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

In which region did this happen?

50

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

In which region did this happen?

51

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

In which country did this happen?

52

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

Where did this happen?

53

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

What caused the event?

54

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

What was the cause of the event?

55

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

What source started the event?

56

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

How was the event first detected?

57

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

How many people were ill?

58

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

How many people were hospitalized?

59

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

How many people were dead?

60

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

Which contaminants or viruses or bacteria were found?

61

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

Which were the symptoms?

62

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

What did the patients have?

63

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

What were the first steps?

64

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

What did they do to control the problem?

65

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

What did the local authorities advise?

66

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

What were the control measures?

67

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

What type of samples were examined?

68

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

What did they test for in the samples?

69

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

What is the concentration of the pathogens?

70

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

What steps were taken to restore the problem?

71

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

What was done to fix the problem?

72

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

What could have been done to prevent the event?

73

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

What could have been done to prevent the event?

74

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

What could have been done to prevent the event?

75

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

How to prevent this?

76

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

What were the investigation steps?

77

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

What did the investigation find?

78

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

How was the infrastructure affected?

79

A community waterborne outbreak of salmonellosis and the effectiveness of a boil water order

Objectives. A 1993 large waterborne outbreak of Salmonella typhimurium infections in Gideon, Mo, a city of 1100 with an unchlorinated community water supply, was investigated to determine the source of contamination and the effectiveness of an order to boil water. Methods A survey of household members in Gideon and the surrounding township produced information on diarrheal illness, water consumption, and compliance with the boil water order. Results More than 650 persons were ill; 15 were hospitalized, and 7 died. Persons consuming city water were more likely to be ill (relative risk [RR] = 9.1, 95% confidence interval [CI] = 2.9, 28.4), and the attack rate increased with increased water consumption. S. typhimurium was recovered from samples taken from a city fire hydrant and a water storage tower. Persons in 318 (30/98) of city households had drunk unboiled water after being informed about the boil water order, including 14 individuals who subsequently became ill. Reasons for noncompliance included ''not remembering'' (44%) and ''disbelieving'' (25%) the order. Conclusions Communities with deteriorating water systems risk widespread illness unless water supplies are properly operated and maintained. Effective education to improve compliance during boil water orders is needed. Introduction Community waterborne disease outbreaks are uncommon in the United States" 2; when they occur, however, many people may be affected.3'4 Community water supplies are drawn from either surface (lakes, rivers, or streams) or deep (groundwater) sources. Since surface waters are easily contaminated, water systems with surface sources are required to be disinfected, usually by chlorination.5 In contrast, deep water sources are considered pure and are not required to be disinfected.5 However, even water drawn from pristine sources may become contaminated during storage or distribution, especially if the water system is not adequately maintained. There are 60 000 community water systems in the United States, serving 91% of the population.2 Two thirds of these systems are not chlorinated, and many have decaying water storage or distribution systems.6 Since several pathogens, including Salmonella,7 Campylobacter, and Cryptosporidium spp,9 can be isolated from the environment or from birds and other animal species, water systems that are inadequately maintained, especially those open to the environment, can become contaminated and cause human illness. Community water systems are routinely monitored for coliform bacteria, and, if contamination is detected, local officials may issue orders for residents to boil water.10'11 However, studies to determine whether boil orders are effective are lacking. 12 In late November 1993, seven culture-confirmed cases of Salmonella typhimurium gastroenteritis among Anderson Township residents were reported to the Missouri Department of Health. Food histories revealed no common food exposures, but all patients had drunk water in Gideon, the only municipality in the township. Gideon municipal water was tested and found to contain fecal coliforms; Gideon residents were warned, via a local radio station, to boil their drinking water, and an investigation was launched to determine the source of the contamination, the magnitude of the outbreak, and the effectiveness of the boil water order. Methods Household Survey Anderson Township, located in southeastern Missouri, is 5 x 10 miles (8 X 16 km) in size and had a 1990 US census population of 1599, of whom 1104 resided in Gideon. A systematic random sample of 150 households was drawn from the 548 households on the Anderson Township personal tax roster. Trained interviewers conducted interviews by telephone or, if no telephone number was available, in person between December 31, 1993, and January 6, 1994. Families not residing in Anderson Township were excluded. All family members 18 years of age or older were asked about the occurrence of diarrheal illness between November 1 and December 31, 1993. Residents of Gideon were also asked about compliance with the order to boil water. A case of acute gastroenteritis was defined as three or more loose stools in a 24-hour period between November 1 and December 31, 1993. Persons with acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 until the day of onset of diarrhea. Persons without acute gastroenteritis were asked about daily water consumption and visits to Gideon from November 15 to November 30, 1993. All families were asked about restaurant or grocery store patronage during November 15 through 30. The t test for clustered samples was used to analyze data from the household survey.13 Relative risks (RRs) with 95% confidence intervals (CIs) were determined. The Mantel-Haenszel test for trend and the Pearson correlation coefficient were used for dose responses. All P values are two-sided. Environmental Studies As required for routine monitoring, Gideon municipal employees collected and submitted one 100-mL water sample twice a month to the Missouri State Public Health Laboratory, which tested the samples for total and fecal coliforms using the membrane filtration method.'4 After the outbreak had been recognized, the Missouri Department of Natural Resources collected additional 100-mL water samples. The Department of Health and the Department of Natural Resources also collected 1-gal (3.8-L) samples from the water distribution system; these samples were filtered and tested for the presence of Salmonella by the Public Health Laboratory via an enrichment technique with tetrathionate and selenite cystine broth.14 On January 5 and 6, the US Environmental Protection Agency collected 1-L samples from throughout the municipal system, including samples from the wells and the city water towers and sediment from the privately owned water tower in Gideon. These samples were tested for Salmonella at the Environmental Protection Agency's Drinking Water Research Division Laboratory in Cincinnati, Ohio.'4 The Environmental Protection Agency also used a computer model to develop dynamic scenarios describing transport of contaminated water through the distribution system.'5 Laboratory Studies Stool or rectal swab specimens were collected from hospitalized patients and from residents of the Gideon nursing home who had diarrhea. Standard methods were used in processing samples, at the clinical laboratories of local hospitals and at the Missouri State Public Health Laboratory, for Salmonella, Shigella, and Campylobacter Serotyping and biochemical testing were performed by the Public Health Laboratory and the Foodbome and Diarrheal Diseases Laboratory of the Centers for Disease Control and Prevention (CDC). Selected environmental and clinical Salmonella isolates were also subtyped via pulsed-field gel electrophoresis at CDC.16 Results In early December 1993, absenteeism at the Gideon schools increased by 250%, "and the sales of antidiarrheal medicines increased by 600%. By the end of December, 15 Anderson Township residents had been hospitalized with S. typhimurium infections, and diarrhea had been reported for 28 of the 68 nursing home residents, 7 of whom died. Household Survey Of the 150 households selected for the survey, 22 were excluded because they were not in Anderson Township. Of the remaining 128 households, 122 were contacted, and 120 households with 329 members participated. Ninety-two (77%) of the participating households were within the city limits of Gideon. Therefore, information was collected on 246 (22%) of 1104 persons living in the city (according to US census figures) and 83 (17%) of 495 outside the city. The age and sex distributions of the participants were similar to 1990 US census figures. One hundred thirty-two participants had diarrhea that met the case definition. The date of onset of the initial cases was November 12, and the peak of onset was November 20 (Figure 1). The median duration of diarrhea was 4 days. Diarrheal illness was accompanied by abdominal cramps (in 68% of cases), self-reported fever (50%), vomiting (42%), and blood in the stool (8%). Of the 132 case patients in the survey, 38 (29%) sought medical attention and 5 (4%) were hospitalized. Persons living inside the city of Gideon (on the municipal water system) were more likely to report having diarrhea than those living outside the city (on private well water) (RR = 1.6, 95% CI = 1.1, 2.3) (Table 1). Among persons living outside the city, those who visited Gideon were more likely to have diarrhea than those who did not visit Gideon during this time (RR = 5.6,95% CI = 1.4, 22.4). Visitors to Gideon who drank Gideon municipal water between November 15 and 30 were also substantially more likely to become ill than those who visited but did not drink Gideon municipal water (RR = 9.1, 95% CI = 2.9, 28.4). No other exposures explained the differences in attack rates seen between those living inside the city and those living outside the city. Gideon residents were not significantly more likely than those living outside the city to shop at any of the four grocery stores in Gideon or neighboring towns or to eat at any of the seven local restaurants between November 15 and 30. In addition, strong dose-response effects were noted among people living in Gideon, with increased consumption of municipal water associated with increased attack rates of diarrhea (P < .01) (Figure 2). We used the results of the household survey to estimate the number of Anderson Township residents who became ill with gastroenteritis during this outbreak. Extrapolating the 44% attack rate for city residents and the 28% attack rate for township residents living outside the city to the 1990 census, we estimate that 650 cases of acute gastroenteritis occurred among Anderson Township residents between November 1 and December 31, 1993. Boil Water Order All but 1 of the 92 Gideon households in the survey were aware of the order to boil water, but 9 (10%) had not heard about the order until an information sheet was delivered on December 28, 10 days after the boil order had been issued. Thirty households (31%) reported that someone in the house had drunk unboiled water after being informed about the order; 12 of 14 (86%) persons who developed diarrhea after the issuance of the boil order reported drinking unboiled water after being informed about the order. The most common reasons for noncompliance were "forgetting" (44%), not believing the initial notification (25%), and not understanding that ice should be made with boiled water (17%). After the delivery of the information sheet, all households surveyed reported believing there was a problem and understanding that ice should be made from boiled water. Environmental Studies The Gideon municipal water system was designed and constructed in the mid-1930s. An average of 130 000 gal of water per day is provided by two adjacent wells 1300 ft (390 m) deep; water was not chlorinated prior to this outbreak. Two city-owned water storage towers (50 000 and 100 000 gal) maintain system pressure; a third privately owned water tower (100 000 gal) was also connected to the system. On November 10, 1993, the municipal water system was flushed by sequentially opening each of the city's 50 fire hydrants for 15 minutes, beginning near the water towers, at an approximate rate of 750 gal per minute (>500 000 gal total). This was the most extensive flushing of the system in at least 3 years. Bimonthly water samples submitted by the city from 1992 and 1993, including samples collected on November 8 and December 6, 1993, showed no evidence of bacterial contamination. However, coliforms were detected in 1 of 3 water samples collected on December 16 (7 coliforms/100 mL), 1 of 7 collected on December 17 (2 coliforms/100 mL), 4 of 6 collected on December 20 (3 to 6 coliforms/100 mL), and 9 of 13 collected on December 21 (6 to 18 coliforms/100 mL). Fecal coliforms were identified in the water samples containing coliforms that were collected on December 16, 20 (all 4 samples), and 21 (3 of 9 samples). One of the six 1-gal samples from the distribution system, collected from a fire hydrant on December 23, grew dulcitolnegative S. typhimurium. In-line automatic chlorination was begun on December 28, 1993; adequate chlorine residuals were obtained by early January 1994. A search was conducted for the possible source of the contamination in the municipal water supply. Contamination of the aquifer was thought to be unlikely because there was no evidence of illness in neighboring towns sharing the same aquifer. In addition, no coliforms were detected from two 1-L water samples collected at the pumps (unchlorinated water obtained directly from the aquifer) on January 5. Given the apparent integrity of the aquifer, attention was focused on the water distribution system and water storage towers. No problems were detected with the distribution system, but several problems were noted with the water towers. Birds were observed frequently roosting on the roofs of the towers, and inspectors found an unscreened overflow pipe and holes 5 in (12.5 cm) in diameter (large enough to allow birds to enter) at the top of the private water storage tank. This tower was subsequently disconnected from the municipal system and drained. Sediment collected from the tower on January 5, 1994, grew dulcitol-negative S. typhimurium. However, the back-flow prevention valve between the private tower and the municipal system was operating properly, indicating that water from the municipal system could enter the privately owned water storage tower but probably not the reverse. The two other water storage towers were inspected on January 12. The 100 000-gal city water storage tower had an improper roof vent and an uncovered hatch, and bird feathers were observed in the tank. Water from the 100 000-gal tower was not obtained for culture because, by that time, the municipal water had been chlorinated. City officials were not aware of any previous inspections of the towers. Results from the Environmental Protection Agency computer model indicated that, during routine operation, most of the water in the distribution system was supplied directly by the wells, while most of the water in the storage towers remained stagnant. The computer model also indicated that the November 10 flushing of the distribution system would have emptied the water towers and that the 100 000-gal tower was the most likely source of the contamination because the first reported illness occurred among patients who lived in the areas that received the greatest amount of water from the 100 000-gal tower. During the first weeks of the outbreak, the attack rates for diarrhea were also higher among people who lived in the areas supplied by the 100 000-gal tower. Laboratory Studies Stool cultures from 15 Anderson Township residents yielded Salmonella; 10 of these strains were serotyped, and each was S. typhimurium that did not ferment dulcitol. Salmonella was also isolated from 13 of 19 stool samples collected from nursing home residents with diarrhea; 12 of these isolates were serotyped, including strains from 4 of the 7 residents who died, and each was dulcitol-negative S. typhimurium. Between September 1, 1993, and January 31, 1994, the Missouri State Public Health Laboratory identified 37 S. typhimurium isolates from persons who resided outside of Anderson Township; 12 (32%) of these sporadic isolates were dulcitol negative. The isolates from the 7 outbreakassociated patients and from the fire hydrant were indistinguishable by pulsedfield gel electrophoresis, and different by one band from the water tower isolate. Isolates from 7 persons with randomly selected sporadic cases identified at the Missouri State Public Health Laboratory were also subtyped by pulsed-field gel electrophoresis; one was different from all others, five had an identical pattem distinct from the outbreak-associated patients, and one was indistinguishable from the outbreak-associated patient isolates. Discussion. We have described a waterbome outbreak of S. typhimurium that apparently caused at least 650 cases of diarrhea, 15 hospitalizations, and 7 deaths in rural Missouri between November 1993 and January 1994. The epidemiologic association between illness and city water consumption is strong, consistent, and dose related. The outbreak was confined to individuals who were exposed to Gideon municipal water. People who drank Gideon municipal water were more likely to become ill, an association that persisted throughout the analysis and became stronger with increased municipal water consumption. Identification of S. typhimurium with an uncommon dulcitol-negative biochemical marker in specimens collected from patients, the water distribution system, and the water storage tower confirmed that the municipal water system was the vehicle of the outbreak. Contamination apparently entered the water system through the 100 000-gal water storage tower, which had an improper roof vent and an uncovered hatch that could allow free access by wild birds. Feathers were discovered in this tank. Birds frequently carry Salmonella spp.6 and have been associated with other waterbome outbreaks of diarrheal disease17-19 and with the suspected contamination of an open-top water treatment tower.17 If the 100 000-gal water storage tower had become contaminated, contaminated water may have been present in the tower for a considerable period since there was minimal mixing between the tower and the distribution system during routine operations. The flushing of more than 500 000 gal of water from the municipal system on November 10 emptied each of the city water storage towers and apparently disseminated the contaminated water into the distribution system. S. typhimurium isolates collected from outbreak-associated patients and from the fire hydrant were indistinguishable by pulsed-field gel electrophoresis, and differed slightly from the isolate collected from the water storage tower. Nevertheless, these isolates may have originated from the same source; the tower isolate differed from the patient and hydrant isolates by only a single band on pulsed-field gel electrophoresis. Chlorine may have altered the tower strain, since chlorine was added to the water towers on December 23 and the isolate from the tower was collected 13 days later. Seven persons died in this outbreak, the most deaths, at that time, associated with a waterbome outbreak in the United States in more than 50 years. The seven patients who died were nursing home residents. Foodbome outbreaks of Salmonella infections in nursing homes are not uncommon,20 and nursing home residents are known to be at a higher risk for serious outcomes of Salmonella infections, including death.21 Although there are no requirements for nursing homes to use chlorinated water, most of the illness associated with this outbreak in the nursing home would have been avoided if the water had been disinfected. Many residents continued to drink unboiled water after the order to boil water had been issued. Many of the people who heard the initial boil order did not appreciate the severity of the situation; the initial boil order gave no reason for its issuance and did not mention associated illness. Only when the information sheets, which clearly explained the rationale and boiling procedure, were delivered to all residences did compliance with the order improve. When boil orders are issued, water supply operators, local govemments, and public health officials should ensure that all residents are adequately informed about the health risks and consequences of noncompliance. Boil orders should be issued with easy-to understand instructions; current recommendations are to bring water to a rolling boil for 1 minute.22 In small towns with limited media outlets, door-to-door delivery of information sheets should begin as early as possible. This outbreak demonstrates the need for proper maintenance of water storage and distribution systems, adequate training in water sanitation for water system operators, and disinfection of all community water systems, including systems that obtain their water from groundwater sources. This outbreak could have been avoided with proper water system maintenance and with adequate ongoing disinfection. The Environmental Protection Agency anticipates the enactment of regulations requiring disinfection of all water supplies; until then, the 40 000 communities with unchlorinated water systems, especially those with poorly operated or deteriorating distribution systems, risk widespread illness from contaminated drinking water.23

What were the infrastructure complaints?

80

A community-wide acute diarrheal disease outbreak associated with drinking contaminated water from shallow bore-wells in a tribal village

Abstract Background: In 2016, India reported 709 acute diarrheal disease (ADD) outbreaks (> 25% of all outbreaks). Tribal populations are at higher risk with 27% not having accessibility to safe drinking water and 75% households not having toilets. On June 26, 2017 Pedda-Gujjul-Thanda, a tribal village reported an acute diarrheal disease (ADD) outbreak. We investigated to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods: We defined a case as ≥3 loose stools within 24 h in Pedda-Gujjul-Thanda residents from June 24–30, 2017. We identified cases by reviewing hospital records and house-to-house survey. We conducted a retrospective cohort study and collected stool samples for culture. We assessed drinking water supply and sanitation practices and tested water samples for faecal-contamination. Results: We identified 191 cases (65% females) with median age 36 years (range 4–80 years) and no deaths. The attack-rate (AR) was 37% (191/512). Downhill colonies (located on slope of hilly terrains of the village) reported higher ARs (56%[136/243], p < 0.001) than others (20%[55/269]). Symptoms included diarrhea (100%), fever (17%), vomiting (16%) and abdominal pain (13%). Drinking water from five shallow bore-wells located in downhill colonies was significantly associated with illness (RR = 4.6, 95%CI = 3.4–6.1 and population attributable fraction 61%). In multivariate analysis, drinking water from the shallow bore-wells located in downhill colonies (aOR = 7.9, [95% CI =4.7– 13.2]), illiteracy (aOR =6, [95% CI = 3.6–10.1]), good hand-washing practice (aOR = 0.4, [95%CI = 0.2–0.7]) and household water treatment (aOR = 0.3, [95%CI = 0.2–0.5]) were significantly associated with illness. Two stool cultures were negative for Vibrio cholerae. Heavy rainfall was reported from June 22–24. Five of six water samples collected from shallow bore-wells located in downhill colonies were positive for faecal contamination. Conclusion: An ADD outbreak with high attack rate in a remote tribal village was associated with drinking water from shallow downhill bore-wells, likely contaminated via runoff from open defecation areas after heavy rains. Based on our recommendations, immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families, and as long-term public health measures construction of house-hold latrines and piped-water supply initiated. Keywords: Acute diarrheal disease, Outbreak, Bore-well, Tribal Introduction Globally there are an estimated 1.7 billion cases and 2.2 million deaths from acute diarrheal disease (ADD) every year [1]. In India, the burden is particularly high with more than 13.9 million cases reported in 2016 and 709 ADD outbreaks reported accounting to more than 25% of all outbreaks [2, 3]. Lack of access to safe drinking water and basic sanitation are the leading causes of ADD burden globally and in India. It is estimated that globally 58% of ADD deaths are attributed to inadequate drinking water, sanitation and hygiene [4]. The WHO/UNICEF Joint Monitoring Program for Water Supply, Sanitation and Hygiene (JMP) 2017 report revealed that 844 million people worldwide lack access to basic drinking-water service and 2.3 billion lack basic sanitation services, while 892 million still practiced open defecation [5]. The National Family Health Survey (NFHS-4, 2015–16) reported that in India only 52% of urban households and 18% of rural households have piped water supply, and the main source of water supply among rural households is bore-wells or tube-wells (51%). It has been estimated that 39% of households in India (54% among rural households) have no toilet facility and practicing open defecation [6]. The “indigenous†populations are socially, culturally and economically isolated and usually lack access to basic drinking-water and sanitation services. Therefore, they are vulnerable to ADD outbreaks and other emerging and re-emerging diseases [7]. The United Nations estimates that there are 370 million indigenous people existing across 90 countries of the world. They constitute 5% of the world population but 15% of the poorest [8]. India alone houses more than 705 such indigenous groups termed as Scheduled Tribes. As per the Census 2011, the total Scheduled Tribe population of India is 10.43 crore with a significant proportion of them living in rural areas [9]. On June 26, 2017, Kama-reddy district of Telangana state reported 55 ADD cases from the Pedda-GujjulThanda village. We conducted the outbreak investigation to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods Setting Pedda-Gujjul-Thanda village is a small tribal village with a total population of 563. The village is remotely located as an isolated community with a hilly terrain and is resource-limited with poor accessibility to sanitation and hygiene facilities. The nearest health care facility available for the residents is located at a distance of 10 km from the village. Case definition We defined a case as three or more loose stools within 24 h in a resident of the Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Case finding To find cases, we reviewed medical records of local health care facilities accessed by village residents in the nearby town. We conducted a medical camp in the village during the outbreak period for five days. We conducted a house-to-house survey in the village to find more cases, which are niether reported to health facility nor medical camp. Retrospective cohort study We conducted a retrospective cohort study to identify risk factors associated with illness. We defined the cohort as residents of Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Village resident was the unit of analysis. For data collection, we trained five teams of local paramedical staff. Using a pre-structured questionnaire, we collected data on demographic characteristics and risk factors related to drinking water, sanitation and hygiene. Good hand-washing practice was defined as reported washing of hands with soap and water every time after defecation and before eating. A bore-well less than 30-m-deep, as assessed from the records of village administration, was considered a shallow bore-well. Laboratory and environmental investigations Two stool samples were collected by the treating physician from admitted patients on the first day of hospital admission and transported to the state reference laboratory within two hours in Cary-Blair transport medium. The samples were cultured for Vibrio cholerae, Salmonella and Shigella on nutrient agar, MacConkey agar and deoxycholate citrate agar. Enteric pathogens were identified by biochemical reaction and by agglutination with anti-sera. We collected details of recent rainfall and conducted an environmental survey with household as sampling unit to assess drinking water, sanitation and hygiene practices. We assessed availability of residual chlorine in all village bore-wells and tested four of five bore-wells in the most affected colonies for faecal contamination by H2S method in field. Water was filled up to the ‘fill line’ of the sample bottle and incubated at room temperature (250–370 C) for 36–48 h and observed for colour change in the medium. A water sample was suspected to be contaminated with faecal matter, if it turned black [10, 11]. Because of limited supplies, we were unable to assess the fifth bore-well. Data analysis We analysed the data to describe the occurrence of cases over time, place, and person. We calculated relative risks (RR) with 95% confidence intervals (CI), population attributable risk percentages and conducted multiple logistic regression analysis with the dependent variables including consumption of shallow-downhill bore-well water, report of visible contaminants like mud in drinking water, illiteracy, household water treatment and good hand-washing practice. We used Epi Info version 7.2 for statistical analysis. Results Descriptive epidemiology We identified 191 ADD cases (65% females), with a village attack rate (AR) of 37% (191/512). The attack rate increased with age, with highest among > 60-year age group (55%) and lowest among children under-10 years (11%) (Table 1). No deaths were reported. In addition to diarrhea, cases presented with fever (17%), vomiting (16%) and abdominal pain (13%). 72% (138/191) cases reported to health care facilities and the medical camp conducted in the village. Among the 191 cases, 159 (83%) had mild illness treated with oral rehydration solution; 30 (16%) had moderate dehydration treated with intravenous fluids on out-patient basis, and 2 (1%) with severe dehydration were admitted in the district hospital for treatment with antibiotics (metronidazole and ciprofloxacin) and intravenous fluids. Cases started reported on June 26, 2017, with onset of symptoms from 24 June 2017. Maximum cases were reported on June 27, 2017, and no new cases were reported after June 30, 2017 (Fig. 1). The tribal population in the village had four sub-tribes namely Katroth, Badhawath, Nenawath, Baromath who resided in seven geographically demarcated colonies (labelled as A to G). Katroth sub-tribe resided in colonies A, B and G; Badhawath in colonies C and D; Nenawath in colony E and Baromath in colony F (Table 2). Colonies B and C had higher attack rates (65 and 47% respectively) as compared to other colonies (Fig. 2). Retrospective cohort study Among 563 village residents, 512 (91%) participated in the study. Among the 512 participants, median age was 28 years (range 1–80 years) with 52% females; 50% reported as illiterate with agriculture as the main source of livelihood for 76%. We analysed possible risk factors associated with ADD (Table 3). Drinking water from bore-well groundwater (vs canned water) was found significantly associated with ADD (RR = 12.7; 95% CI = 1.8–87.4). However, only 32 (6%) residents in the village used canned water and bore-well groundwater was the predominant source of water supply. Therefore, we analysed the water sources further, by location and type of bore-wells. Residents who used any of the five shallow bore-wells located downhill were significantly at higher risk (RR = 4.6; 95% CI = 3.4–6.1) and deep bore-wells were protective (RR = 0.4; 95% CI = 0.2–0.9). Report of visible contaminants like mud in drinking water (aOR = 4; 95% CI = 2.1–7.6) and illiteracy (aOR = 3.6; 95% CI = 3.5–10.1) were significantly associated with illness; and household water treatment (done either by boiling or use of candle filters) (aOR = 0.4; 95% CI = 0.2–0.7) and good hand-washing practice (aOR = 0.2; 95% CI = 0.1–0.5) were found protective. Laboratory and environmental results Stool samples collected from two hospitalized cases showed no growth for Vibrio cholerae, Salmonella and Shigella on culture. Among 110 households, 100 (91%) were available for environmental survey. Among the 100 houses surveyed, 79 (79%) were kutcha (low quality) type, made of mud, thatch and other low-quality material. Only 5 (5%) households had a designated toilet at home while the remaining 95 (95%) practiced open defecation at a site located on the slope of the hill behind the downhill colonies B and C (Figs. 2 and 3). Bore-wells were the main source of drinking water supply for 93 (93%) households. There were two deep borewells provided by the village administration and 17 shallow type bore-wells privately constructed by village residents. Five of these 17 (30%) shallow bore-wells were located in colonies B and C, on the downhill slope below the open defecation site. Plastic pipelines from the shallow wells were improperly installed with leakages at multiple points. There was no facility at source, for chlorination or any other mode of purification. Thirty households (30%) treated the water before consumption either by boiling or by use of candle filters. There was no routine drinking water surveillance in place by any authority for assessing the quality and fitness for drinking water. There was no residual chlorine found in any water samples. Three of four drinking water samples from bore-wells of most affected colonies (B and C) indicated faecal contamination by H2S field testing. There was heavy rainfall (average 65 mm in a day) from 22 to 24 June 2017. Prevention and control measures undertaken to contain the outbreak The village residents were discouraged from using shallow bore-well water and were provided with safe canned drinking water until all leakages were secured. Leakages in water supply from the bore-wells were identified and secured. Chlorine tablets were distributed for household level water disinfection. We informed the residents to avoid open defecation near drinking water sources and residential premises. Public health staff conducted health education daily to improve awareness among the villagers about water, sanitation, and hygiene. After active implementation of these control measures, cases declined rapidly in the village (Fig. 1). Discussion A rapid systematic epidemiological investigation of this outbreak identified water contamination points and likely mode of contamination. Based on these findings and our recommendations, the local health department instituted immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families. Effective implementation of public health measures limited the exposure of the community to contaminated water source resulting in rapid containment of the outbreak. Waterborne disease outbreaks tend to have cases spread over a time-period due to ongoing exposure to the contaminated water. In contrast, the pattern of epidemic curve in the present waterborne disease outbreak resembled that of food-borne with a point source exposure. Heavy rains contributed to the run-off of water from the open defecation site into the ground water of shallow wells located on slopes of hilly terrain resulting in heavy contamination and sudden rise of cases. Rapid control measures in the small village, implemented effectively within a short period of time, may have led to rapid decline of cases. The available epidemiological evidence also did not support generation of hypothesis of food-borne origin of the outbreak. In an outbreak reported among school children in Northern Greece in 2012, investigation revealed a waterborne viral gastroenteritis outbreak with a point source pattern, due to consumption of heavily contaminated water from a tap, which was not in use for two weeks during Christmas vacation [12]. Attack rate was high in this outbreak (37%), possibly due to exposure to high pathogen load subsequent to gross faecal contamination of water sources. In the absence of other alternative water sources, this tribal community was exclusively dependent on the contaminated water source for drinking, therefore exposing a large section of the community to risk. Geetha et al. analysed 32 diarrheal outbreaks in south India in non- tribal communities and reported lower attack rates varying from 0.6 to 21.5% [13]. However, tribal populations in India such as in Pedda-Gujjul-Thanda are marginalized with poor availability of WASH facilities [14]. This vulnerable tribal population continues to be at higher risk for ADD outbreaks with 27% not having access to safe drinking water and 75% of households not having toilets [15]. They need special assistance schemes from the government to enable them overcome poor accessibility to WASH facilities and secure healthy living [16]. Due to inadequate availability of communally managed safe public water points by the local authority, this community in Pedda-Gujjul-Thanda village was dependent on privately constructed shallow bore-wells for water supply. These are economical but likely to be unsafe. In this outbreak, open defecation site was present on the downhill slope in proximity to the residential premises and water resources, increasing the risk of drinking water contamination. Among the entire village population, 61% of ADD cases were attributable to drinking water from the ‘shallow downhill bore-wells’ (Population Attributable Fraction 61%), which was also evident from rapid outbreak containment following the elimination of exposure to this single risk factor. Since this exposure factor is amenable to long-term public health intervention, permanent elimination of shallow downhill borewells as water source was recommended, replacing them with properly secured deep bore-wells. Shallow bore-wells are known for their susceptibility to contamination from surface land-use activities [17, 18]. Studies have found levels of E. coli and enteric viruses to be high in shallow sources of ground water especially when they are in close proximity to polluting sources [19–21]. Consumption of ground water from shallow bore-wells with no purification facility increases the risk of diarrhea outbreaks manifold [22, 23]. A metaanalytic study of water-borne diarrheal disease outbreaks in China reported that 78 of 85 (92%) outbreaks (between year 1987 to 2014) were due to poor sanitary conditions of wells with lavatories/septic tanks nearby and lack of purification facilities [24]. In developed countries and urban areas of developing countries, as water supply and sanitation have improved dramatically over a period of time, such outbreaks were rarely reported in the recent past. The largest E. coli O157 outbreak in United States occurred in 1999 at a county fair (781 ill persons and 2 deaths) was due to groundwater source from a temporary unregulated well at the fairground [25]. Our findings have implication for India’s progress towards United Nation’s Sustainable Development Goal (SDG) 6 and India’s nation-wide campaign ‘Swachh Bharat Mission (SBM)’ to ensure availability and management of water and sanitation for all. SDG 6 aims at achieving universal access to basic sanitation service by 2030; and it has been reported that between 2000 and 2017, the proportion lacking even a basic sanitation service decreased from 44 to 27% [26]. SBM aims to achieve an open-defecation free status in rural areas through the construction of household-owned and community-owned toilets and establishing an accountable mechanism of monitoring toilet use. In 2015 in India, around 524 million (39%) practiced open defecation. However, under the SBM mission, due to increase in ‘households with toilets’ only 19 million (1.4%) practiced open defecation in January 2019 [5, 27]. There has also been a 71.58% increase in ‘households with toilets’ from October 2014 to October 2019 in rural areas of the Telangana state in India [27]. The tribal community initially obstructed the effective delivery of health care services; however, after involvement of the local stakeholders and tribal leaders, the acceptance towards medical treatment and community health services improved. Notwithstanding, most of the patients were still reluctant and did not consent for giving stool specimens for laboratory diagnosis. Establishing a rapport with the reticent tribal community was a major challenge faced by the outbreak investigation team. Lack of microbiological aetiology confirmation of the outbreak remained a limitation of the investigation due to limited stool samples and laboratory-capacity constraints of the remote area. Recognizing the pivotal importance of SDGs, national health policy of India (2017) has set the health-related cross-sectorial goal “access to safe water and sanitation to all by 2020†[28]. Greater political and financial commitment towards resource-limited remote tribal areas with effective community mobilization is required to accelerate the public health interventions to improve WASH and to prevent ADD outbreaks in the future. Conclusion This was a community-wide acute diarrheal disease outbreak with high village attack rate in a remote tribal village of Telangana with poor availability of safe water, sanitation and hygiene (WASH) facilities. A rapid and systematic epidemiological investigation identified drinking of faecal-contaminated water from the shallow borewells as the leading cause for this outbreak. These borewells were likely contaminated from runoff after rain from open defecation areas located on a downhill slope. Prompt and targeted public health action contained the number of cases.

What happened?

81

A community-wide acute diarrheal disease outbreak associated with drinking contaminated water from shallow bore-wells in a tribal village

Abstract Background: In 2016, India reported 709 acute diarrheal disease (ADD) outbreaks (> 25% of all outbreaks). Tribal populations are at higher risk with 27% not having accessibility to safe drinking water and 75% households not having toilets. On June 26, 2017 Pedda-Gujjul-Thanda, a tribal village reported an acute diarrheal disease (ADD) outbreak. We investigated to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods: We defined a case as ≥3 loose stools within 24 h in Pedda-Gujjul-Thanda residents from June 24–30, 2017. We identified cases by reviewing hospital records and house-to-house survey. We conducted a retrospective cohort study and collected stool samples for culture. We assessed drinking water supply and sanitation practices and tested water samples for faecal-contamination. Results: We identified 191 cases (65% females) with median age 36 years (range 4–80 years) and no deaths. The attack-rate (AR) was 37% (191/512). Downhill colonies (located on slope of hilly terrains of the village) reported higher ARs (56%[136/243], p < 0.001) than others (20%[55/269]). Symptoms included diarrhea (100%), fever (17%), vomiting (16%) and abdominal pain (13%). Drinking water from five shallow bore-wells located in downhill colonies was significantly associated with illness (RR = 4.6, 95%CI = 3.4–6.1 and population attributable fraction 61%). In multivariate analysis, drinking water from the shallow bore-wells located in downhill colonies (aOR = 7.9, [95% CI =4.7– 13.2]), illiteracy (aOR =6, [95% CI = 3.6–10.1]), good hand-washing practice (aOR = 0.4, [95%CI = 0.2–0.7]) and household water treatment (aOR = 0.3, [95%CI = 0.2–0.5]) were significantly associated with illness. Two stool cultures were negative for Vibrio cholerae. Heavy rainfall was reported from June 22–24. Five of six water samples collected from shallow bore-wells located in downhill colonies were positive for faecal contamination. Conclusion: An ADD outbreak with high attack rate in a remote tribal village was associated with drinking water from shallow downhill bore-wells, likely contaminated via runoff from open defecation areas after heavy rains. Based on our recommendations, immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families, and as long-term public health measures construction of house-hold latrines and piped-water supply initiated. Keywords: Acute diarrheal disease, Outbreak, Bore-well, Tribal Introduction Globally there are an estimated 1.7 billion cases and 2.2 million deaths from acute diarrheal disease (ADD) every year [1]. In India, the burden is particularly high with more than 13.9 million cases reported in 2016 and 709 ADD outbreaks reported accounting to more than 25% of all outbreaks [2, 3]. Lack of access to safe drinking water and basic sanitation are the leading causes of ADD burden globally and in India. It is estimated that globally 58% of ADD deaths are attributed to inadequate drinking water, sanitation and hygiene [4]. The WHO/UNICEF Joint Monitoring Program for Water Supply, Sanitation and Hygiene (JMP) 2017 report revealed that 844 million people worldwide lack access to basic drinking-water service and 2.3 billion lack basic sanitation services, while 892 million still practiced open defecation [5]. The National Family Health Survey (NFHS-4, 2015–16) reported that in India only 52% of urban households and 18% of rural households have piped water supply, and the main source of water supply among rural households is bore-wells or tube-wells (51%). It has been estimated that 39% of households in India (54% among rural households) have no toilet facility and practicing open defecation [6]. The “indigenous†populations are socially, culturally and economically isolated and usually lack access to basic drinking-water and sanitation services. Therefore, they are vulnerable to ADD outbreaks and other emerging and re-emerging diseases [7]. The United Nations estimates that there are 370 million indigenous people existing across 90 countries of the world. They constitute 5% of the world population but 15% of the poorest [8]. India alone houses more than 705 such indigenous groups termed as Scheduled Tribes. As per the Census 2011, the total Scheduled Tribe population of India is 10.43 crore with a significant proportion of them living in rural areas [9]. On June 26, 2017, Kama-reddy district of Telangana state reported 55 ADD cases from the Pedda-GujjulThanda village. We conducted the outbreak investigation to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods Setting Pedda-Gujjul-Thanda village is a small tribal village with a total population of 563. The village is remotely located as an isolated community with a hilly terrain and is resource-limited with poor accessibility to sanitation and hygiene facilities. The nearest health care facility available for the residents is located at a distance of 10 km from the village. Case definition We defined a case as three or more loose stools within 24 h in a resident of the Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Case finding To find cases, we reviewed medical records of local health care facilities accessed by village residents in the nearby town. We conducted a medical camp in the village during the outbreak period for five days. We conducted a house-to-house survey in the village to find more cases, which are niether reported to health facility nor medical camp. Retrospective cohort study We conducted a retrospective cohort study to identify risk factors associated with illness. We defined the cohort as residents of Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Village resident was the unit of analysis. For data collection, we trained five teams of local paramedical staff. Using a pre-structured questionnaire, we collected data on demographic characteristics and risk factors related to drinking water, sanitation and hygiene. Good hand-washing practice was defined as reported washing of hands with soap and water every time after defecation and before eating. A bore-well less than 30-m-deep, as assessed from the records of village administration, was considered a shallow bore-well. Laboratory and environmental investigations Two stool samples were collected by the treating physician from admitted patients on the first day of hospital admission and transported to the state reference laboratory within two hours in Cary-Blair transport medium. The samples were cultured for Vibrio cholerae, Salmonella and Shigella on nutrient agar, MacConkey agar and deoxycholate citrate agar. Enteric pathogens were identified by biochemical reaction and by agglutination with anti-sera. We collected details of recent rainfall and conducted an environmental survey with household as sampling unit to assess drinking water, sanitation and hygiene practices. We assessed availability of residual chlorine in all village bore-wells and tested four of five bore-wells in the most affected colonies for faecal contamination by H2S method in field. Water was filled up to the ‘fill line’ of the sample bottle and incubated at room temperature (250–370 C) for 36–48 h and observed for colour change in the medium. A water sample was suspected to be contaminated with faecal matter, if it turned black [10, 11]. Because of limited supplies, we were unable to assess the fifth bore-well. Data analysis We analysed the data to describe the occurrence of cases over time, place, and person. We calculated relative risks (RR) with 95% confidence intervals (CI), population attributable risk percentages and conducted multiple logistic regression analysis with the dependent variables including consumption of shallow-downhill bore-well water, report of visible contaminants like mud in drinking water, illiteracy, household water treatment and good hand-washing practice. We used Epi Info version 7.2 for statistical analysis. Results Descriptive epidemiology We identified 191 ADD cases (65% females), with a village attack rate (AR) of 37% (191/512). The attack rate increased with age, with highest among > 60-year age group (55%) and lowest among children under-10 years (11%) (Table 1). No deaths were reported. In addition to diarrhea, cases presented with fever (17%), vomiting (16%) and abdominal pain (13%). 72% (138/191) cases reported to health care facilities and the medical camp conducted in the village. Among the 191 cases, 159 (83%) had mild illness treated with oral rehydration solution; 30 (16%) had moderate dehydration treated with intravenous fluids on out-patient basis, and 2 (1%) with severe dehydration were admitted in the district hospital for treatment with antibiotics (metronidazole and ciprofloxacin) and intravenous fluids. Cases started reported on June 26, 2017, with onset of symptoms from 24 June 2017. Maximum cases were reported on June 27, 2017, and no new cases were reported after June 30, 2017 (Fig. 1). The tribal population in the village had four sub-tribes namely Katroth, Badhawath, Nenawath, Baromath who resided in seven geographically demarcated colonies (labelled as A to G). Katroth sub-tribe resided in colonies A, B and G; Badhawath in colonies C and D; Nenawath in colony E and Baromath in colony F (Table 2). Colonies B and C had higher attack rates (65 and 47% respectively) as compared to other colonies (Fig. 2). Retrospective cohort study Among 563 village residents, 512 (91%) participated in the study. Among the 512 participants, median age was 28 years (range 1–80 years) with 52% females; 50% reported as illiterate with agriculture as the main source of livelihood for 76%. We analysed possible risk factors associated with ADD (Table 3). Drinking water from bore-well groundwater (vs canned water) was found significantly associated with ADD (RR = 12.7; 95% CI = 1.8–87.4). However, only 32 (6%) residents in the village used canned water and bore-well groundwater was the predominant source of water supply. Therefore, we analysed the water sources further, by location and type of bore-wells. Residents who used any of the five shallow bore-wells located downhill were significantly at higher risk (RR = 4.6; 95% CI = 3.4–6.1) and deep bore-wells were protective (RR = 0.4; 95% CI = 0.2–0.9). Report of visible contaminants like mud in drinking water (aOR = 4; 95% CI = 2.1–7.6) and illiteracy (aOR = 3.6; 95% CI = 3.5–10.1) were significantly associated with illness; and household water treatment (done either by boiling or use of candle filters) (aOR = 0.4; 95% CI = 0.2–0.7) and good hand-washing practice (aOR = 0.2; 95% CI = 0.1–0.5) were found protective. Laboratory and environmental results Stool samples collected from two hospitalized cases showed no growth for Vibrio cholerae, Salmonella and Shigella on culture. Among 110 households, 100 (91%) were available for environmental survey. Among the 100 houses surveyed, 79 (79%) were kutcha (low quality) type, made of mud, thatch and other low-quality material. Only 5 (5%) households had a designated toilet at home while the remaining 95 (95%) practiced open defecation at a site located on the slope of the hill behind the downhill colonies B and C (Figs. 2 and 3). Bore-wells were the main source of drinking water supply for 93 (93%) households. There were two deep borewells provided by the village administration and 17 shallow type bore-wells privately constructed by village residents. Five of these 17 (30%) shallow bore-wells were located in colonies B and C, on the downhill slope below the open defecation site. Plastic pipelines from the shallow wells were improperly installed with leakages at multiple points. There was no facility at source, for chlorination or any other mode of purification. Thirty households (30%) treated the water before consumption either by boiling or by use of candle filters. There was no routine drinking water surveillance in place by any authority for assessing the quality and fitness for drinking water. There was no residual chlorine found in any water samples. Three of four drinking water samples from bore-wells of most affected colonies (B and C) indicated faecal contamination by H2S field testing. There was heavy rainfall (average 65 mm in a day) from 22 to 24 June 2017. Prevention and control measures undertaken to contain the outbreak The village residents were discouraged from using shallow bore-well water and were provided with safe canned drinking water until all leakages were secured. Leakages in water supply from the bore-wells were identified and secured. Chlorine tablets were distributed for household level water disinfection. We informed the residents to avoid open defecation near drinking water sources and residential premises. Public health staff conducted health education daily to improve awareness among the villagers about water, sanitation, and hygiene. After active implementation of these control measures, cases declined rapidly in the village (Fig. 1). Discussion A rapid systematic epidemiological investigation of this outbreak identified water contamination points and likely mode of contamination. Based on these findings and our recommendations, the local health department instituted immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families. Effective implementation of public health measures limited the exposure of the community to contaminated water source resulting in rapid containment of the outbreak. Waterborne disease outbreaks tend to have cases spread over a time-period due to ongoing exposure to the contaminated water. In contrast, the pattern of epidemic curve in the present waterborne disease outbreak resembled that of food-borne with a point source exposure. Heavy rains contributed to the run-off of water from the open defecation site into the ground water of shallow wells located on slopes of hilly terrain resulting in heavy contamination and sudden rise of cases. Rapid control measures in the small village, implemented effectively within a short period of time, may have led to rapid decline of cases. The available epidemiological evidence also did not support generation of hypothesis of food-borne origin of the outbreak. In an outbreak reported among school children in Northern Greece in 2012, investigation revealed a waterborne viral gastroenteritis outbreak with a point source pattern, due to consumption of heavily contaminated water from a tap, which was not in use for two weeks during Christmas vacation [12]. Attack rate was high in this outbreak (37%), possibly due to exposure to high pathogen load subsequent to gross faecal contamination of water sources. In the absence of other alternative water sources, this tribal community was exclusively dependent on the contaminated water source for drinking, therefore exposing a large section of the community to risk. Geetha et al. analysed 32 diarrheal outbreaks in south India in non- tribal communities and reported lower attack rates varying from 0.6 to 21.5% [13]. However, tribal populations in India such as in Pedda-Gujjul-Thanda are marginalized with poor availability of WASH facilities [14]. This vulnerable tribal population continues to be at higher risk for ADD outbreaks with 27% not having access to safe drinking water and 75% of households not having toilets [15]. They need special assistance schemes from the government to enable them overcome poor accessibility to WASH facilities and secure healthy living [16]. Due to inadequate availability of communally managed safe public water points by the local authority, this community in Pedda-Gujjul-Thanda village was dependent on privately constructed shallow bore-wells for water supply. These are economical but likely to be unsafe. In this outbreak, open defecation site was present on the downhill slope in proximity to the residential premises and water resources, increasing the risk of drinking water contamination. Among the entire village population, 61% of ADD cases were attributable to drinking water from the ‘shallow downhill bore-wells’ (Population Attributable Fraction 61%), which was also evident from rapid outbreak containment following the elimination of exposure to this single risk factor. Since this exposure factor is amenable to long-term public health intervention, permanent elimination of shallow downhill borewells as water source was recommended, replacing them with properly secured deep bore-wells. Shallow bore-wells are known for their susceptibility to contamination from surface land-use activities [17, 18]. Studies have found levels of E. coli and enteric viruses to be high in shallow sources of ground water especially when they are in close proximity to polluting sources [19–21]. Consumption of ground water from shallow bore-wells with no purification facility increases the risk of diarrhea outbreaks manifold [22, 23]. A metaanalytic study of water-borne diarrheal disease outbreaks in China reported that 78 of 85 (92%) outbreaks (between year 1987 to 2014) were due to poor sanitary conditions of wells with lavatories/septic tanks nearby and lack of purification facilities [24]. In developed countries and urban areas of developing countries, as water supply and sanitation have improved dramatically over a period of time, such outbreaks were rarely reported in the recent past. The largest E. coli O157 outbreak in United States occurred in 1999 at a county fair (781 ill persons and 2 deaths) was due to groundwater source from a temporary unregulated well at the fairground [25]. Our findings have implication for India’s progress towards United Nation’s Sustainable Development Goal (SDG) 6 and India’s nation-wide campaign ‘Swachh Bharat Mission (SBM)’ to ensure availability and management of water and sanitation for all. SDG 6 aims at achieving universal access to basic sanitation service by 2030; and it has been reported that between 2000 and 2017, the proportion lacking even a basic sanitation service decreased from 44 to 27% [26]. SBM aims to achieve an open-defecation free status in rural areas through the construction of household-owned and community-owned toilets and establishing an accountable mechanism of monitoring toilet use. In 2015 in India, around 524 million (39%) practiced open defecation. However, under the SBM mission, due to increase in ‘households with toilets’ only 19 million (1.4%) practiced open defecation in January 2019 [5, 27]. There has also been a 71.58% increase in ‘households with toilets’ from October 2014 to October 2019 in rural areas of the Telangana state in India [27]. The tribal community initially obstructed the effective delivery of health care services; however, after involvement of the local stakeholders and tribal leaders, the acceptance towards medical treatment and community health services improved. Notwithstanding, most of the patients were still reluctant and did not consent for giving stool specimens for laboratory diagnosis. Establishing a rapport with the reticent tribal community was a major challenge faced by the outbreak investigation team. Lack of microbiological aetiology confirmation of the outbreak remained a limitation of the investigation due to limited stool samples and laboratory-capacity constraints of the remote area. Recognizing the pivotal importance of SDGs, national health policy of India (2017) has set the health-related cross-sectorial goal “access to safe water and sanitation to all by 2020†[28]. Greater political and financial commitment towards resource-limited remote tribal areas with effective community mobilization is required to accelerate the public health interventions to improve WASH and to prevent ADD outbreaks in the future. Conclusion This was a community-wide acute diarrheal disease outbreak with high village attack rate in a remote tribal village of Telangana with poor availability of safe water, sanitation and hygiene (WASH) facilities. A rapid and systematic epidemiological investigation identified drinking of faecal-contaminated water from the shallow borewells as the leading cause for this outbreak. These borewells were likely contaminated from runoff after rain from open defecation areas located on a downhill slope. Prompt and targeted public health action contained the number of cases.

What was the event?

82

A community-wide acute diarrheal disease outbreak associated with drinking contaminated water from shallow bore-wells in a tribal village

Abstract Background: In 2016, India reported 709 acute diarrheal disease (ADD) outbreaks (> 25% of all outbreaks). Tribal populations are at higher risk with 27% not having accessibility to safe drinking water and 75% households not having toilets. On June 26, 2017 Pedda-Gujjul-Thanda, a tribal village reported an acute diarrheal disease (ADD) outbreak. We investigated to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods: We defined a case as ≥3 loose stools within 24 h in Pedda-Gujjul-Thanda residents from June 24–30, 2017. We identified cases by reviewing hospital records and house-to-house survey. We conducted a retrospective cohort study and collected stool samples for culture. We assessed drinking water supply and sanitation practices and tested water samples for faecal-contamination. Results: We identified 191 cases (65% females) with median age 36 years (range 4–80 years) and no deaths. The attack-rate (AR) was 37% (191/512). Downhill colonies (located on slope of hilly terrains of the village) reported higher ARs (56%[136/243], p < 0.001) than others (20%[55/269]). Symptoms included diarrhea (100%), fever (17%), vomiting (16%) and abdominal pain (13%). Drinking water from five shallow bore-wells located in downhill colonies was significantly associated with illness (RR = 4.6, 95%CI = 3.4–6.1 and population attributable fraction 61%). In multivariate analysis, drinking water from the shallow bore-wells located in downhill colonies (aOR = 7.9, [95% CI =4.7– 13.2]), illiteracy (aOR =6, [95% CI = 3.6–10.1]), good hand-washing practice (aOR = 0.4, [95%CI = 0.2–0.7]) and household water treatment (aOR = 0.3, [95%CI = 0.2–0.5]) were significantly associated with illness. Two stool cultures were negative for Vibrio cholerae. Heavy rainfall was reported from June 22–24. Five of six water samples collected from shallow bore-wells located in downhill colonies were positive for faecal contamination. Conclusion: An ADD outbreak with high attack rate in a remote tribal village was associated with drinking water from shallow downhill bore-wells, likely contaminated via runoff from open defecation areas after heavy rains. Based on our recommendations, immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families, and as long-term public health measures construction of house-hold latrines and piped-water supply initiated. Keywords: Acute diarrheal disease, Outbreak, Bore-well, Tribal Introduction Globally there are an estimated 1.7 billion cases and 2.2 million deaths from acute diarrheal disease (ADD) every year [1]. In India, the burden is particularly high with more than 13.9 million cases reported in 2016 and 709 ADD outbreaks reported accounting to more than 25% of all outbreaks [2, 3]. Lack of access to safe drinking water and basic sanitation are the leading causes of ADD burden globally and in India. It is estimated that globally 58% of ADD deaths are attributed to inadequate drinking water, sanitation and hygiene [4]. The WHO/UNICEF Joint Monitoring Program for Water Supply, Sanitation and Hygiene (JMP) 2017 report revealed that 844 million people worldwide lack access to basic drinking-water service and 2.3 billion lack basic sanitation services, while 892 million still practiced open defecation [5]. The National Family Health Survey (NFHS-4, 2015–16) reported that in India only 52% of urban households and 18% of rural households have piped water supply, and the main source of water supply among rural households is bore-wells or tube-wells (51%). It has been estimated that 39% of households in India (54% among rural households) have no toilet facility and practicing open defecation [6]. The “indigenous†populations are socially, culturally and economically isolated and usually lack access to basic drinking-water and sanitation services. Therefore, they are vulnerable to ADD outbreaks and other emerging and re-emerging diseases [7]. The United Nations estimates that there are 370 million indigenous people existing across 90 countries of the world. They constitute 5% of the world population but 15% of the poorest [8]. India alone houses more than 705 such indigenous groups termed as Scheduled Tribes. As per the Census 2011, the total Scheduled Tribe population of India is 10.43 crore with a significant proportion of them living in rural areas [9]. On June 26, 2017, Kama-reddy district of Telangana state reported 55 ADD cases from the Pedda-GujjulThanda village. We conducted the outbreak investigation to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods Setting Pedda-Gujjul-Thanda village is a small tribal village with a total population of 563. The village is remotely located as an isolated community with a hilly terrain and is resource-limited with poor accessibility to sanitation and hygiene facilities. The nearest health care facility available for the residents is located at a distance of 10 km from the village. Case definition We defined a case as three or more loose stools within 24 h in a resident of the Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Case finding To find cases, we reviewed medical records of local health care facilities accessed by village residents in the nearby town. We conducted a medical camp in the village during the outbreak period for five days. We conducted a house-to-house survey in the village to find more cases, which are niether reported to health facility nor medical camp. Retrospective cohort study We conducted a retrospective cohort study to identify risk factors associated with illness. We defined the cohort as residents of Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Village resident was the unit of analysis. For data collection, we trained five teams of local paramedical staff. Using a pre-structured questionnaire, we collected data on demographic characteristics and risk factors related to drinking water, sanitation and hygiene. Good hand-washing practice was defined as reported washing of hands with soap and water every time after defecation and before eating. A bore-well less than 30-m-deep, as assessed from the records of village administration, was considered a shallow bore-well. Laboratory and environmental investigations Two stool samples were collected by the treating physician from admitted patients on the first day of hospital admission and transported to the state reference laboratory within two hours in Cary-Blair transport medium. The samples were cultured for Vibrio cholerae, Salmonella and Shigella on nutrient agar, MacConkey agar and deoxycholate citrate agar. Enteric pathogens were identified by biochemical reaction and by agglutination with anti-sera. We collected details of recent rainfall and conducted an environmental survey with household as sampling unit to assess drinking water, sanitation and hygiene practices. We assessed availability of residual chlorine in all village bore-wells and tested four of five bore-wells in the most affected colonies for faecal contamination by H2S method in field. Water was filled up to the ‘fill line’ of the sample bottle and incubated at room temperature (250–370 C) for 36–48 h and observed for colour change in the medium. A water sample was suspected to be contaminated with faecal matter, if it turned black [10, 11]. Because of limited supplies, we were unable to assess the fifth bore-well. Data analysis We analysed the data to describe the occurrence of cases over time, place, and person. We calculated relative risks (RR) with 95% confidence intervals (CI), population attributable risk percentages and conducted multiple logistic regression analysis with the dependent variables including consumption of shallow-downhill bore-well water, report of visible contaminants like mud in drinking water, illiteracy, household water treatment and good hand-washing practice. We used Epi Info version 7.2 for statistical analysis. Results Descriptive epidemiology We identified 191 ADD cases (65% females), with a village attack rate (AR) of 37% (191/512). The attack rate increased with age, with highest among > 60-year age group (55%) and lowest among children under-10 years (11%) (Table 1). No deaths were reported. In addition to diarrhea, cases presented with fever (17%), vomiting (16%) and abdominal pain (13%). 72% (138/191) cases reported to health care facilities and the medical camp conducted in the village. Among the 191 cases, 159 (83%) had mild illness treated with oral rehydration solution; 30 (16%) had moderate dehydration treated with intravenous fluids on out-patient basis, and 2 (1%) with severe dehydration were admitted in the district hospital for treatment with antibiotics (metronidazole and ciprofloxacin) and intravenous fluids. Cases started reported on June 26, 2017, with onset of symptoms from 24 June 2017. Maximum cases were reported on June 27, 2017, and no new cases were reported after June 30, 2017 (Fig. 1). The tribal population in the village had four sub-tribes namely Katroth, Badhawath, Nenawath, Baromath who resided in seven geographically demarcated colonies (labelled as A to G). Katroth sub-tribe resided in colonies A, B and G; Badhawath in colonies C and D; Nenawath in colony E and Baromath in colony F (Table 2). Colonies B and C had higher attack rates (65 and 47% respectively) as compared to other colonies (Fig. 2). Retrospective cohort study Among 563 village residents, 512 (91%) participated in the study. Among the 512 participants, median age was 28 years (range 1–80 years) with 52% females; 50% reported as illiterate with agriculture as the main source of livelihood for 76%. We analysed possible risk factors associated with ADD (Table 3). Drinking water from bore-well groundwater (vs canned water) was found significantly associated with ADD (RR = 12.7; 95% CI = 1.8–87.4). However, only 32 (6%) residents in the village used canned water and bore-well groundwater was the predominant source of water supply. Therefore, we analysed the water sources further, by location and type of bore-wells. Residents who used any of the five shallow bore-wells located downhill were significantly at higher risk (RR = 4.6; 95% CI = 3.4–6.1) and deep bore-wells were protective (RR = 0.4; 95% CI = 0.2–0.9). Report of visible contaminants like mud in drinking water (aOR = 4; 95% CI = 2.1–7.6) and illiteracy (aOR = 3.6; 95% CI = 3.5–10.1) were significantly associated with illness; and household water treatment (done either by boiling or use of candle filters) (aOR = 0.4; 95% CI = 0.2–0.7) and good hand-washing practice (aOR = 0.2; 95% CI = 0.1–0.5) were found protective. Laboratory and environmental results Stool samples collected from two hospitalized cases showed no growth for Vibrio cholerae, Salmonella and Shigella on culture. Among 110 households, 100 (91%) were available for environmental survey. Among the 100 houses surveyed, 79 (79%) were kutcha (low quality) type, made of mud, thatch and other low-quality material. Only 5 (5%) households had a designated toilet at home while the remaining 95 (95%) practiced open defecation at a site located on the slope of the hill behind the downhill colonies B and C (Figs. 2 and 3). Bore-wells were the main source of drinking water supply for 93 (93%) households. There were two deep borewells provided by the village administration and 17 shallow type bore-wells privately constructed by village residents. Five of these 17 (30%) shallow bore-wells were located in colonies B and C, on the downhill slope below the open defecation site. Plastic pipelines from the shallow wells were improperly installed with leakages at multiple points. There was no facility at source, for chlorination or any other mode of purification. Thirty households (30%) treated the water before consumption either by boiling or by use of candle filters. There was no routine drinking water surveillance in place by any authority for assessing the quality and fitness for drinking water. There was no residual chlorine found in any water samples. Three of four drinking water samples from bore-wells of most affected colonies (B and C) indicated faecal contamination by H2S field testing. There was heavy rainfall (average 65 mm in a day) from 22 to 24 June 2017. Prevention and control measures undertaken to contain the outbreak The village residents were discouraged from using shallow bore-well water and were provided with safe canned drinking water until all leakages were secured. Leakages in water supply from the bore-wells were identified and secured. Chlorine tablets were distributed for household level water disinfection. We informed the residents to avoid open defecation near drinking water sources and residential premises. Public health staff conducted health education daily to improve awareness among the villagers about water, sanitation, and hygiene. After active implementation of these control measures, cases declined rapidly in the village (Fig. 1). Discussion A rapid systematic epidemiological investigation of this outbreak identified water contamination points and likely mode of contamination. Based on these findings and our recommendations, the local health department instituted immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families. Effective implementation of public health measures limited the exposure of the community to contaminated water source resulting in rapid containment of the outbreak. Waterborne disease outbreaks tend to have cases spread over a time-period due to ongoing exposure to the contaminated water. In contrast, the pattern of epidemic curve in the present waterborne disease outbreak resembled that of food-borne with a point source exposure. Heavy rains contributed to the run-off of water from the open defecation site into the ground water of shallow wells located on slopes of hilly terrain resulting in heavy contamination and sudden rise of cases. Rapid control measures in the small village, implemented effectively within a short period of time, may have led to rapid decline of cases. The available epidemiological evidence also did not support generation of hypothesis of food-borne origin of the outbreak. In an outbreak reported among school children in Northern Greece in 2012, investigation revealed a waterborne viral gastroenteritis outbreak with a point source pattern, due to consumption of heavily contaminated water from a tap, which was not in use for two weeks during Christmas vacation [12]. Attack rate was high in this outbreak (37%), possibly due to exposure to high pathogen load subsequent to gross faecal contamination of water sources. In the absence of other alternative water sources, this tribal community was exclusively dependent on the contaminated water source for drinking, therefore exposing a large section of the community to risk. Geetha et al. analysed 32 diarrheal outbreaks in south India in non- tribal communities and reported lower attack rates varying from 0.6 to 21.5% [13]. However, tribal populations in India such as in Pedda-Gujjul-Thanda are marginalized with poor availability of WASH facilities [14]. This vulnerable tribal population continues to be at higher risk for ADD outbreaks with 27% not having access to safe drinking water and 75% of households not having toilets [15]. They need special assistance schemes from the government to enable them overcome poor accessibility to WASH facilities and secure healthy living [16]. Due to inadequate availability of communally managed safe public water points by the local authority, this community in Pedda-Gujjul-Thanda village was dependent on privately constructed shallow bore-wells for water supply. These are economical but likely to be unsafe. In this outbreak, open defecation site was present on the downhill slope in proximity to the residential premises and water resources, increasing the risk of drinking water contamination. Among the entire village population, 61% of ADD cases were attributable to drinking water from the ‘shallow downhill bore-wells’ (Population Attributable Fraction 61%), which was also evident from rapid outbreak containment following the elimination of exposure to this single risk factor. Since this exposure factor is amenable to long-term public health intervention, permanent elimination of shallow downhill borewells as water source was recommended, replacing them with properly secured deep bore-wells. Shallow bore-wells are known for their susceptibility to contamination from surface land-use activities [17, 18]. Studies have found levels of E. coli and enteric viruses to be high in shallow sources of ground water especially when they are in close proximity to polluting sources [19–21]. Consumption of ground water from shallow bore-wells with no purification facility increases the risk of diarrhea outbreaks manifold [22, 23]. A metaanalytic study of water-borne diarrheal disease outbreaks in China reported that 78 of 85 (92%) outbreaks (between year 1987 to 2014) were due to poor sanitary conditions of wells with lavatories/septic tanks nearby and lack of purification facilities [24]. In developed countries and urban areas of developing countries, as water supply and sanitation have improved dramatically over a period of time, such outbreaks were rarely reported in the recent past. The largest E. coli O157 outbreak in United States occurred in 1999 at a county fair (781 ill persons and 2 deaths) was due to groundwater source from a temporary unregulated well at the fairground [25]. Our findings have implication for India’s progress towards United Nation’s Sustainable Development Goal (SDG) 6 and India’s nation-wide campaign ‘Swachh Bharat Mission (SBM)’ to ensure availability and management of water and sanitation for all. SDG 6 aims at achieving universal access to basic sanitation service by 2030; and it has been reported that between 2000 and 2017, the proportion lacking even a basic sanitation service decreased from 44 to 27% [26]. SBM aims to achieve an open-defecation free status in rural areas through the construction of household-owned and community-owned toilets and establishing an accountable mechanism of monitoring toilet use. In 2015 in India, around 524 million (39%) practiced open defecation. However, under the SBM mission, due to increase in ‘households with toilets’ only 19 million (1.4%) practiced open defecation in January 2019 [5, 27]. There has also been a 71.58% increase in ‘households with toilets’ from October 2014 to October 2019 in rural areas of the Telangana state in India [27]. The tribal community initially obstructed the effective delivery of health care services; however, after involvement of the local stakeholders and tribal leaders, the acceptance towards medical treatment and community health services improved. Notwithstanding, most of the patients were still reluctant and did not consent for giving stool specimens for laboratory diagnosis. Establishing a rapport with the reticent tribal community was a major challenge faced by the outbreak investigation team. Lack of microbiological aetiology confirmation of the outbreak remained a limitation of the investigation due to limited stool samples and laboratory-capacity constraints of the remote area. Recognizing the pivotal importance of SDGs, national health policy of India (2017) has set the health-related cross-sectorial goal “access to safe water and sanitation to all by 2020†[28]. Greater political and financial commitment towards resource-limited remote tribal areas with effective community mobilization is required to accelerate the public health interventions to improve WASH and to prevent ADD outbreaks in the future. Conclusion This was a community-wide acute diarrheal disease outbreak with high village attack rate in a remote tribal village of Telangana with poor availability of safe water, sanitation and hygiene (WASH) facilities. A rapid and systematic epidemiological investigation identified drinking of faecal-contaminated water from the shallow borewells as the leading cause for this outbreak. These borewells were likely contaminated from runoff after rain from open defecation areas located on a downhill slope. Prompt and targeted public health action contained the number of cases.

When did this happen?

83

A community-wide acute diarrheal disease outbreak associated with drinking contaminated water from shallow bore-wells in a tribal village

Abstract Background: In 2016, India reported 709 acute diarrheal disease (ADD) outbreaks (> 25% of all outbreaks). Tribal populations are at higher risk with 27% not having accessibility to safe drinking water and 75% households not having toilets. On June 26, 2017 Pedda-Gujjul-Thanda, a tribal village reported an acute diarrheal disease (ADD) outbreak. We investigated to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods: We defined a case as ≥3 loose stools within 24 h in Pedda-Gujjul-Thanda residents from June 24–30, 2017. We identified cases by reviewing hospital records and house-to-house survey. We conducted a retrospective cohort study and collected stool samples for culture. We assessed drinking water supply and sanitation practices and tested water samples for faecal-contamination. Results: We identified 191 cases (65% females) with median age 36 years (range 4–80 years) and no deaths. The attack-rate (AR) was 37% (191/512). Downhill colonies (located on slope of hilly terrains of the village) reported higher ARs (56%[136/243], p < 0.001) than others (20%[55/269]). Symptoms included diarrhea (100%), fever (17%), vomiting (16%) and abdominal pain (13%). Drinking water from five shallow bore-wells located in downhill colonies was significantly associated with illness (RR = 4.6, 95%CI = 3.4–6.1 and population attributable fraction 61%). In multivariate analysis, drinking water from the shallow bore-wells located in downhill colonies (aOR = 7.9, [95% CI =4.7– 13.2]), illiteracy (aOR =6, [95% CI = 3.6–10.1]), good hand-washing practice (aOR = 0.4, [95%CI = 0.2–0.7]) and household water treatment (aOR = 0.3, [95%CI = 0.2–0.5]) were significantly associated with illness. Two stool cultures were negative for Vibrio cholerae. Heavy rainfall was reported from June 22–24. Five of six water samples collected from shallow bore-wells located in downhill colonies were positive for faecal contamination. Conclusion: An ADD outbreak with high attack rate in a remote tribal village was associated with drinking water from shallow downhill bore-wells, likely contaminated via runoff from open defecation areas after heavy rains. Based on our recommendations, immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families, and as long-term public health measures construction of house-hold latrines and piped-water supply initiated. Keywords: Acute diarrheal disease, Outbreak, Bore-well, Tribal Introduction Globally there are an estimated 1.7 billion cases and 2.2 million deaths from acute diarrheal disease (ADD) every year [1]. In India, the burden is particularly high with more than 13.9 million cases reported in 2016 and 709 ADD outbreaks reported accounting to more than 25% of all outbreaks [2, 3]. Lack of access to safe drinking water and basic sanitation are the leading causes of ADD burden globally and in India. It is estimated that globally 58% of ADD deaths are attributed to inadequate drinking water, sanitation and hygiene [4]. The WHO/UNICEF Joint Monitoring Program for Water Supply, Sanitation and Hygiene (JMP) 2017 report revealed that 844 million people worldwide lack access to basic drinking-water service and 2.3 billion lack basic sanitation services, while 892 million still practiced open defecation [5]. The National Family Health Survey (NFHS-4, 2015–16) reported that in India only 52% of urban households and 18% of rural households have piped water supply, and the main source of water supply among rural households is bore-wells or tube-wells (51%). It has been estimated that 39% of households in India (54% among rural households) have no toilet facility and practicing open defecation [6]. The “indigenous†populations are socially, culturally and economically isolated and usually lack access to basic drinking-water and sanitation services. Therefore, they are vulnerable to ADD outbreaks and other emerging and re-emerging diseases [7]. The United Nations estimates that there are 370 million indigenous people existing across 90 countries of the world. They constitute 5% of the world population but 15% of the poorest [8]. India alone houses more than 705 such indigenous groups termed as Scheduled Tribes. As per the Census 2011, the total Scheduled Tribe population of India is 10.43 crore with a significant proportion of them living in rural areas [9]. On June 26, 2017, Kama-reddy district of Telangana state reported 55 ADD cases from the Pedda-GujjulThanda village. We conducted the outbreak investigation to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods Setting Pedda-Gujjul-Thanda village is a small tribal village with a total population of 563. The village is remotely located as an isolated community with a hilly terrain and is resource-limited with poor accessibility to sanitation and hygiene facilities. The nearest health care facility available for the residents is located at a distance of 10 km from the village. Case definition We defined a case as three or more loose stools within 24 h in a resident of the Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Case finding To find cases, we reviewed medical records of local health care facilities accessed by village residents in the nearby town. We conducted a medical camp in the village during the outbreak period for five days. We conducted a house-to-house survey in the village to find more cases, which are niether reported to health facility nor medical camp. Retrospective cohort study We conducted a retrospective cohort study to identify risk factors associated with illness. We defined the cohort as residents of Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Village resident was the unit of analysis. For data collection, we trained five teams of local paramedical staff. Using a pre-structured questionnaire, we collected data on demographic characteristics and risk factors related to drinking water, sanitation and hygiene. Good hand-washing practice was defined as reported washing of hands with soap and water every time after defecation and before eating. A bore-well less than 30-m-deep, as assessed from the records of village administration, was considered a shallow bore-well. Laboratory and environmental investigations Two stool samples were collected by the treating physician from admitted patients on the first day of hospital admission and transported to the state reference laboratory within two hours in Cary-Blair transport medium. The samples were cultured for Vibrio cholerae, Salmonella and Shigella on nutrient agar, MacConkey agar and deoxycholate citrate agar. Enteric pathogens were identified by biochemical reaction and by agglutination with anti-sera. We collected details of recent rainfall and conducted an environmental survey with household as sampling unit to assess drinking water, sanitation and hygiene practices. We assessed availability of residual chlorine in all village bore-wells and tested four of five bore-wells in the most affected colonies for faecal contamination by H2S method in field. Water was filled up to the ‘fill line’ of the sample bottle and incubated at room temperature (250–370 C) for 36–48 h and observed for colour change in the medium. A water sample was suspected to be contaminated with faecal matter, if it turned black [10, 11]. Because of limited supplies, we were unable to assess the fifth bore-well. Data analysis We analysed the data to describe the occurrence of cases over time, place, and person. We calculated relative risks (RR) with 95% confidence intervals (CI), population attributable risk percentages and conducted multiple logistic regression analysis with the dependent variables including consumption of shallow-downhill bore-well water, report of visible contaminants like mud in drinking water, illiteracy, household water treatment and good hand-washing practice. We used Epi Info version 7.2 for statistical analysis. Results Descriptive epidemiology We identified 191 ADD cases (65% females), with a village attack rate (AR) of 37% (191/512). The attack rate increased with age, with highest among > 60-year age group (55%) and lowest among children under-10 years (11%) (Table 1). No deaths were reported. In addition to diarrhea, cases presented with fever (17%), vomiting (16%) and abdominal pain (13%). 72% (138/191) cases reported to health care facilities and the medical camp conducted in the village. Among the 191 cases, 159 (83%) had mild illness treated with oral rehydration solution; 30 (16%) had moderate dehydration treated with intravenous fluids on out-patient basis, and 2 (1%) with severe dehydration were admitted in the district hospital for treatment with antibiotics (metronidazole and ciprofloxacin) and intravenous fluids. Cases started reported on June 26, 2017, with onset of symptoms from 24 June 2017. Maximum cases were reported on June 27, 2017, and no new cases were reported after June 30, 2017 (Fig. 1). The tribal population in the village had four sub-tribes namely Katroth, Badhawath, Nenawath, Baromath who resided in seven geographically demarcated colonies (labelled as A to G). Katroth sub-tribe resided in colonies A, B and G; Badhawath in colonies C and D; Nenawath in colony E and Baromath in colony F (Table 2). Colonies B and C had higher attack rates (65 and 47% respectively) as compared to other colonies (Fig. 2). Retrospective cohort study Among 563 village residents, 512 (91%) participated in the study. Among the 512 participants, median age was 28 years (range 1–80 years) with 52% females; 50% reported as illiterate with agriculture as the main source of livelihood for 76%. We analysed possible risk factors associated with ADD (Table 3). Drinking water from bore-well groundwater (vs canned water) was found significantly associated with ADD (RR = 12.7; 95% CI = 1.8–87.4). However, only 32 (6%) residents in the village used canned water and bore-well groundwater was the predominant source of water supply. Therefore, we analysed the water sources further, by location and type of bore-wells. Residents who used any of the five shallow bore-wells located downhill were significantly at higher risk (RR = 4.6; 95% CI = 3.4–6.1) and deep bore-wells were protective (RR = 0.4; 95% CI = 0.2–0.9). Report of visible contaminants like mud in drinking water (aOR = 4; 95% CI = 2.1–7.6) and illiteracy (aOR = 3.6; 95% CI = 3.5–10.1) were significantly associated with illness; and household water treatment (done either by boiling or use of candle filters) (aOR = 0.4; 95% CI = 0.2–0.7) and good hand-washing practice (aOR = 0.2; 95% CI = 0.1–0.5) were found protective. Laboratory and environmental results Stool samples collected from two hospitalized cases showed no growth for Vibrio cholerae, Salmonella and Shigella on culture. Among 110 households, 100 (91%) were available for environmental survey. Among the 100 houses surveyed, 79 (79%) were kutcha (low quality) type, made of mud, thatch and other low-quality material. Only 5 (5%) households had a designated toilet at home while the remaining 95 (95%) practiced open defecation at a site located on the slope of the hill behind the downhill colonies B and C (Figs. 2 and 3). Bore-wells were the main source of drinking water supply for 93 (93%) households. There were two deep borewells provided by the village administration and 17 shallow type bore-wells privately constructed by village residents. Five of these 17 (30%) shallow bore-wells were located in colonies B and C, on the downhill slope below the open defecation site. Plastic pipelines from the shallow wells were improperly installed with leakages at multiple points. There was no facility at source, for chlorination or any other mode of purification. Thirty households (30%) treated the water before consumption either by boiling or by use of candle filters. There was no routine drinking water surveillance in place by any authority for assessing the quality and fitness for drinking water. There was no residual chlorine found in any water samples. Three of four drinking water samples from bore-wells of most affected colonies (B and C) indicated faecal contamination by H2S field testing. There was heavy rainfall (average 65 mm in a day) from 22 to 24 June 2017. Prevention and control measures undertaken to contain the outbreak The village residents were discouraged from using shallow bore-well water and were provided with safe canned drinking water until all leakages were secured. Leakages in water supply from the bore-wells were identified and secured. Chlorine tablets were distributed for household level water disinfection. We informed the residents to avoid open defecation near drinking water sources and residential premises. Public health staff conducted health education daily to improve awareness among the villagers about water, sanitation, and hygiene. After active implementation of these control measures, cases declined rapidly in the village (Fig. 1). Discussion A rapid systematic epidemiological investigation of this outbreak identified water contamination points and likely mode of contamination. Based on these findings and our recommendations, the local health department instituted immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families. Effective implementation of public health measures limited the exposure of the community to contaminated water source resulting in rapid containment of the outbreak. Waterborne disease outbreaks tend to have cases spread over a time-period due to ongoing exposure to the contaminated water. In contrast, the pattern of epidemic curve in the present waterborne disease outbreak resembled that of food-borne with a point source exposure. Heavy rains contributed to the run-off of water from the open defecation site into the ground water of shallow wells located on slopes of hilly terrain resulting in heavy contamination and sudden rise of cases. Rapid control measures in the small village, implemented effectively within a short period of time, may have led to rapid decline of cases. The available epidemiological evidence also did not support generation of hypothesis of food-borne origin of the outbreak. In an outbreak reported among school children in Northern Greece in 2012, investigation revealed a waterborne viral gastroenteritis outbreak with a point source pattern, due to consumption of heavily contaminated water from a tap, which was not in use for two weeks during Christmas vacation [12]. Attack rate was high in this outbreak (37%), possibly due to exposure to high pathogen load subsequent to gross faecal contamination of water sources. In the absence of other alternative water sources, this tribal community was exclusively dependent on the contaminated water source for drinking, therefore exposing a large section of the community to risk. Geetha et al. analysed 32 diarrheal outbreaks in south India in non- tribal communities and reported lower attack rates varying from 0.6 to 21.5% [13]. However, tribal populations in India such as in Pedda-Gujjul-Thanda are marginalized with poor availability of WASH facilities [14]. This vulnerable tribal population continues to be at higher risk for ADD outbreaks with 27% not having access to safe drinking water and 75% of households not having toilets [15]. They need special assistance schemes from the government to enable them overcome poor accessibility to WASH facilities and secure healthy living [16]. Due to inadequate availability of communally managed safe public water points by the local authority, this community in Pedda-Gujjul-Thanda village was dependent on privately constructed shallow bore-wells for water supply. These are economical but likely to be unsafe. In this outbreak, open defecation site was present on the downhill slope in proximity to the residential premises and water resources, increasing the risk of drinking water contamination. Among the entire village population, 61% of ADD cases were attributable to drinking water from the ‘shallow downhill bore-wells’ (Population Attributable Fraction 61%), which was also evident from rapid outbreak containment following the elimination of exposure to this single risk factor. Since this exposure factor is amenable to long-term public health intervention, permanent elimination of shallow downhill borewells as water source was recommended, replacing them with properly secured deep bore-wells. Shallow bore-wells are known for their susceptibility to contamination from surface land-use activities [17, 18]. Studies have found levels of E. coli and enteric viruses to be high in shallow sources of ground water especially when they are in close proximity to polluting sources [19–21]. Consumption of ground water from shallow bore-wells with no purification facility increases the risk of diarrhea outbreaks manifold [22, 23]. A metaanalytic study of water-borne diarrheal disease outbreaks in China reported that 78 of 85 (92%) outbreaks (between year 1987 to 2014) were due to poor sanitary conditions of wells with lavatories/septic tanks nearby and lack of purification facilities [24]. In developed countries and urban areas of developing countries, as water supply and sanitation have improved dramatically over a period of time, such outbreaks were rarely reported in the recent past. The largest E. coli O157 outbreak in United States occurred in 1999 at a county fair (781 ill persons and 2 deaths) was due to groundwater source from a temporary unregulated well at the fairground [25]. Our findings have implication for India’s progress towards United Nation’s Sustainable Development Goal (SDG) 6 and India’s nation-wide campaign ‘Swachh Bharat Mission (SBM)’ to ensure availability and management of water and sanitation for all. SDG 6 aims at achieving universal access to basic sanitation service by 2030; and it has been reported that between 2000 and 2017, the proportion lacking even a basic sanitation service decreased from 44 to 27% [26]. SBM aims to achieve an open-defecation free status in rural areas through the construction of household-owned and community-owned toilets and establishing an accountable mechanism of monitoring toilet use. In 2015 in India, around 524 million (39%) practiced open defecation. However, under the SBM mission, due to increase in ‘households with toilets’ only 19 million (1.4%) practiced open defecation in January 2019 [5, 27]. There has also been a 71.58% increase in ‘households with toilets’ from October 2014 to October 2019 in rural areas of the Telangana state in India [27]. The tribal community initially obstructed the effective delivery of health care services; however, after involvement of the local stakeholders and tribal leaders, the acceptance towards medical treatment and community health services improved. Notwithstanding, most of the patients were still reluctant and did not consent for giving stool specimens for laboratory diagnosis. Establishing a rapport with the reticent tribal community was a major challenge faced by the outbreak investigation team. Lack of microbiological aetiology confirmation of the outbreak remained a limitation of the investigation due to limited stool samples and laboratory-capacity constraints of the remote area. Recognizing the pivotal importance of SDGs, national health policy of India (2017) has set the health-related cross-sectorial goal “access to safe water and sanitation to all by 2020†[28]. Greater political and financial commitment towards resource-limited remote tribal areas with effective community mobilization is required to accelerate the public health interventions to improve WASH and to prevent ADD outbreaks in the future. Conclusion This was a community-wide acute diarrheal disease outbreak with high village attack rate in a remote tribal village of Telangana with poor availability of safe water, sanitation and hygiene (WASH) facilities. A rapid and systematic epidemiological investigation identified drinking of faecal-contaminated water from the shallow borewells as the leading cause for this outbreak. These borewells were likely contaminated from runoff after rain from open defecation areas located on a downhill slope. Prompt and targeted public health action contained the number of cases.

When did this event start?

84

A community-wide acute diarrheal disease outbreak associated with drinking contaminated water from shallow bore-wells in a tribal village

Abstract Background: In 2016, India reported 709 acute diarrheal disease (ADD) outbreaks (> 25% of all outbreaks). Tribal populations are at higher risk with 27% not having accessibility to safe drinking water and 75% households not having toilets. On June 26, 2017 Pedda-Gujjul-Thanda, a tribal village reported an acute diarrheal disease (ADD) outbreak. We investigated to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods: We defined a case as ≥3 loose stools within 24 h in Pedda-Gujjul-Thanda residents from June 24–30, 2017. We identified cases by reviewing hospital records and house-to-house survey. We conducted a retrospective cohort study and collected stool samples for culture. We assessed drinking water supply and sanitation practices and tested water samples for faecal-contamination. Results: We identified 191 cases (65% females) with median age 36 years (range 4–80 years) and no deaths. The attack-rate (AR) was 37% (191/512). Downhill colonies (located on slope of hilly terrains of the village) reported higher ARs (56%[136/243], p < 0.001) than others (20%[55/269]). Symptoms included diarrhea (100%), fever (17%), vomiting (16%) and abdominal pain (13%). Drinking water from five shallow bore-wells located in downhill colonies was significantly associated with illness (RR = 4.6, 95%CI = 3.4–6.1 and population attributable fraction 61%). In multivariate analysis, drinking water from the shallow bore-wells located in downhill colonies (aOR = 7.9, [95% CI =4.7– 13.2]), illiteracy (aOR =6, [95% CI = 3.6–10.1]), good hand-washing practice (aOR = 0.4, [95%CI = 0.2–0.7]) and household water treatment (aOR = 0.3, [95%CI = 0.2–0.5]) were significantly associated with illness. Two stool cultures were negative for Vibrio cholerae. Heavy rainfall was reported from June 22–24. Five of six water samples collected from shallow bore-wells located in downhill colonies were positive for faecal contamination. Conclusion: An ADD outbreak with high attack rate in a remote tribal village was associated with drinking water from shallow downhill bore-wells, likely contaminated via runoff from open defecation areas after heavy rains. Based on our recommendations, immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families, and as long-term public health measures construction of house-hold latrines and piped-water supply initiated. Keywords: Acute diarrheal disease, Outbreak, Bore-well, Tribal Introduction Globally there are an estimated 1.7 billion cases and 2.2 million deaths from acute diarrheal disease (ADD) every year [1]. In India, the burden is particularly high with more than 13.9 million cases reported in 2016 and 709 ADD outbreaks reported accounting to more than 25% of all outbreaks [2, 3]. Lack of access to safe drinking water and basic sanitation are the leading causes of ADD burden globally and in India. It is estimated that globally 58% of ADD deaths are attributed to inadequate drinking water, sanitation and hygiene [4]. The WHO/UNICEF Joint Monitoring Program for Water Supply, Sanitation and Hygiene (JMP) 2017 report revealed that 844 million people worldwide lack access to basic drinking-water service and 2.3 billion lack basic sanitation services, while 892 million still practiced open defecation [5]. The National Family Health Survey (NFHS-4, 2015–16) reported that in India only 52% of urban households and 18% of rural households have piped water supply, and the main source of water supply among rural households is bore-wells or tube-wells (51%). It has been estimated that 39% of households in India (54% among rural households) have no toilet facility and practicing open defecation [6]. The “indigenous†populations are socially, culturally and economically isolated and usually lack access to basic drinking-water and sanitation services. Therefore, they are vulnerable to ADD outbreaks and other emerging and re-emerging diseases [7]. The United Nations estimates that there are 370 million indigenous people existing across 90 countries of the world. They constitute 5% of the world population but 15% of the poorest [8]. India alone houses more than 705 such indigenous groups termed as Scheduled Tribes. As per the Census 2011, the total Scheduled Tribe population of India is 10.43 crore with a significant proportion of them living in rural areas [9]. On June 26, 2017, Kama-reddy district of Telangana state reported 55 ADD cases from the Pedda-GujjulThanda village. We conducted the outbreak investigation to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods Setting Pedda-Gujjul-Thanda village is a small tribal village with a total population of 563. The village is remotely located as an isolated community with a hilly terrain and is resource-limited with poor accessibility to sanitation and hygiene facilities. The nearest health care facility available for the residents is located at a distance of 10 km from the village. Case definition We defined a case as three or more loose stools within 24 h in a resident of the Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Case finding To find cases, we reviewed medical records of local health care facilities accessed by village residents in the nearby town. We conducted a medical camp in the village during the outbreak period for five days. We conducted a house-to-house survey in the village to find more cases, which are niether reported to health facility nor medical camp. Retrospective cohort study We conducted a retrospective cohort study to identify risk factors associated with illness. We defined the cohort as residents of Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Village resident was the unit of analysis. For data collection, we trained five teams of local paramedical staff. Using a pre-structured questionnaire, we collected data on demographic characteristics and risk factors related to drinking water, sanitation and hygiene. Good hand-washing practice was defined as reported washing of hands with soap and water every time after defecation and before eating. A bore-well less than 30-m-deep, as assessed from the records of village administration, was considered a shallow bore-well. Laboratory and environmental investigations Two stool samples were collected by the treating physician from admitted patients on the first day of hospital admission and transported to the state reference laboratory within two hours in Cary-Blair transport medium. The samples were cultured for Vibrio cholerae, Salmonella and Shigella on nutrient agar, MacConkey agar and deoxycholate citrate agar. Enteric pathogens were identified by biochemical reaction and by agglutination with anti-sera. We collected details of recent rainfall and conducted an environmental survey with household as sampling unit to assess drinking water, sanitation and hygiene practices. We assessed availability of residual chlorine in all village bore-wells and tested four of five bore-wells in the most affected colonies for faecal contamination by H2S method in field. Water was filled up to the ‘fill line’ of the sample bottle and incubated at room temperature (250–370 C) for 36–48 h and observed for colour change in the medium. A water sample was suspected to be contaminated with faecal matter, if it turned black [10, 11]. Because of limited supplies, we were unable to assess the fifth bore-well. Data analysis We analysed the data to describe the occurrence of cases over time, place, and person. We calculated relative risks (RR) with 95% confidence intervals (CI), population attributable risk percentages and conducted multiple logistic regression analysis with the dependent variables including consumption of shallow-downhill bore-well water, report of visible contaminants like mud in drinking water, illiteracy, household water treatment and good hand-washing practice. We used Epi Info version 7.2 for statistical analysis. Results Descriptive epidemiology We identified 191 ADD cases (65% females), with a village attack rate (AR) of 37% (191/512). The attack rate increased with age, with highest among > 60-year age group (55%) and lowest among children under-10 years (11%) (Table 1). No deaths were reported. In addition to diarrhea, cases presented with fever (17%), vomiting (16%) and abdominal pain (13%). 72% (138/191) cases reported to health care facilities and the medical camp conducted in the village. Among the 191 cases, 159 (83%) had mild illness treated with oral rehydration solution; 30 (16%) had moderate dehydration treated with intravenous fluids on out-patient basis, and 2 (1%) with severe dehydration were admitted in the district hospital for treatment with antibiotics (metronidazole and ciprofloxacin) and intravenous fluids. Cases started reported on June 26, 2017, with onset of symptoms from 24 June 2017. Maximum cases were reported on June 27, 2017, and no new cases were reported after June 30, 2017 (Fig. 1). The tribal population in the village had four sub-tribes namely Katroth, Badhawath, Nenawath, Baromath who resided in seven geographically demarcated colonies (labelled as A to G). Katroth sub-tribe resided in colonies A, B and G; Badhawath in colonies C and D; Nenawath in colony E and Baromath in colony F (Table 2). Colonies B and C had higher attack rates (65 and 47% respectively) as compared to other colonies (Fig. 2). Retrospective cohort study Among 563 village residents, 512 (91%) participated in the study. Among the 512 participants, median age was 28 years (range 1–80 years) with 52% females; 50% reported as illiterate with agriculture as the main source of livelihood for 76%. We analysed possible risk factors associated with ADD (Table 3). Drinking water from bore-well groundwater (vs canned water) was found significantly associated with ADD (RR = 12.7; 95% CI = 1.8–87.4). However, only 32 (6%) residents in the village used canned water and bore-well groundwater was the predominant source of water supply. Therefore, we analysed the water sources further, by location and type of bore-wells. Residents who used any of the five shallow bore-wells located downhill were significantly at higher risk (RR = 4.6; 95% CI = 3.4–6.1) and deep bore-wells were protective (RR = 0.4; 95% CI = 0.2–0.9). Report of visible contaminants like mud in drinking water (aOR = 4; 95% CI = 2.1–7.6) and illiteracy (aOR = 3.6; 95% CI = 3.5–10.1) were significantly associated with illness; and household water treatment (done either by boiling or use of candle filters) (aOR = 0.4; 95% CI = 0.2–0.7) and good hand-washing practice (aOR = 0.2; 95% CI = 0.1–0.5) were found protective. Laboratory and environmental results Stool samples collected from two hospitalized cases showed no growth for Vibrio cholerae, Salmonella and Shigella on culture. Among 110 households, 100 (91%) were available for environmental survey. Among the 100 houses surveyed, 79 (79%) were kutcha (low quality) type, made of mud, thatch and other low-quality material. Only 5 (5%) households had a designated toilet at home while the remaining 95 (95%) practiced open defecation at a site located on the slope of the hill behind the downhill colonies B and C (Figs. 2 and 3). Bore-wells were the main source of drinking water supply for 93 (93%) households. There were two deep borewells provided by the village administration and 17 shallow type bore-wells privately constructed by village residents. Five of these 17 (30%) shallow bore-wells were located in colonies B and C, on the downhill slope below the open defecation site. Plastic pipelines from the shallow wells were improperly installed with leakages at multiple points. There was no facility at source, for chlorination or any other mode of purification. Thirty households (30%) treated the water before consumption either by boiling or by use of candle filters. There was no routine drinking water surveillance in place by any authority for assessing the quality and fitness for drinking water. There was no residual chlorine found in any water samples. Three of four drinking water samples from bore-wells of most affected colonies (B and C) indicated faecal contamination by H2S field testing. There was heavy rainfall (average 65 mm in a day) from 22 to 24 June 2017. Prevention and control measures undertaken to contain the outbreak The village residents were discouraged from using shallow bore-well water and were provided with safe canned drinking water until all leakages were secured. Leakages in water supply from the bore-wells were identified and secured. Chlorine tablets were distributed for household level water disinfection. We informed the residents to avoid open defecation near drinking water sources and residential premises. Public health staff conducted health education daily to improve awareness among the villagers about water, sanitation, and hygiene. After active implementation of these control measures, cases declined rapidly in the village (Fig. 1). Discussion A rapid systematic epidemiological investigation of this outbreak identified water contamination points and likely mode of contamination. Based on these findings and our recommendations, the local health department instituted immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families. Effective implementation of public health measures limited the exposure of the community to contaminated water source resulting in rapid containment of the outbreak. Waterborne disease outbreaks tend to have cases spread over a time-period due to ongoing exposure to the contaminated water. In contrast, the pattern of epidemic curve in the present waterborne disease outbreak resembled that of food-borne with a point source exposure. Heavy rains contributed to the run-off of water from the open defecation site into the ground water of shallow wells located on slopes of hilly terrain resulting in heavy contamination and sudden rise of cases. Rapid control measures in the small village, implemented effectively within a short period of time, may have led to rapid decline of cases. The available epidemiological evidence also did not support generation of hypothesis of food-borne origin of the outbreak. In an outbreak reported among school children in Northern Greece in 2012, investigation revealed a waterborne viral gastroenteritis outbreak with a point source pattern, due to consumption of heavily contaminated water from a tap, which was not in use for two weeks during Christmas vacation [12]. Attack rate was high in this outbreak (37%), possibly due to exposure to high pathogen load subsequent to gross faecal contamination of water sources. In the absence of other alternative water sources, this tribal community was exclusively dependent on the contaminated water source for drinking, therefore exposing a large section of the community to risk. Geetha et al. analysed 32 diarrheal outbreaks in south India in non- tribal communities and reported lower attack rates varying from 0.6 to 21.5% [13]. However, tribal populations in India such as in Pedda-Gujjul-Thanda are marginalized with poor availability of WASH facilities [14]. This vulnerable tribal population continues to be at higher risk for ADD outbreaks with 27% not having access to safe drinking water and 75% of households not having toilets [15]. They need special assistance schemes from the government to enable them overcome poor accessibility to WASH facilities and secure healthy living [16]. Due to inadequate availability of communally managed safe public water points by the local authority, this community in Pedda-Gujjul-Thanda village was dependent on privately constructed shallow bore-wells for water supply. These are economical but likely to be unsafe. In this outbreak, open defecation site was present on the downhill slope in proximity to the residential premises and water resources, increasing the risk of drinking water contamination. Among the entire village population, 61% of ADD cases were attributable to drinking water from the ‘shallow downhill bore-wells’ (Population Attributable Fraction 61%), which was also evident from rapid outbreak containment following the elimination of exposure to this single risk factor. Since this exposure factor is amenable to long-term public health intervention, permanent elimination of shallow downhill borewells as water source was recommended, replacing them with properly secured deep bore-wells. Shallow bore-wells are known for their susceptibility to contamination from surface land-use activities [17, 18]. Studies have found levels of E. coli and enteric viruses to be high in shallow sources of ground water especially when they are in close proximity to polluting sources [19–21]. Consumption of ground water from shallow bore-wells with no purification facility increases the risk of diarrhea outbreaks manifold [22, 23]. A metaanalytic study of water-borne diarrheal disease outbreaks in China reported that 78 of 85 (92%) outbreaks (between year 1987 to 2014) were due to poor sanitary conditions of wells with lavatories/septic tanks nearby and lack of purification facilities [24]. In developed countries and urban areas of developing countries, as water supply and sanitation have improved dramatically over a period of time, such outbreaks were rarely reported in the recent past. The largest E. coli O157 outbreak in United States occurred in 1999 at a county fair (781 ill persons and 2 deaths) was due to groundwater source from a temporary unregulated well at the fairground [25]. Our findings have implication for India’s progress towards United Nation’s Sustainable Development Goal (SDG) 6 and India’s nation-wide campaign ‘Swachh Bharat Mission (SBM)’ to ensure availability and management of water and sanitation for all. SDG 6 aims at achieving universal access to basic sanitation service by 2030; and it has been reported that between 2000 and 2017, the proportion lacking even a basic sanitation service decreased from 44 to 27% [26]. SBM aims to achieve an open-defecation free status in rural areas through the construction of household-owned and community-owned toilets and establishing an accountable mechanism of monitoring toilet use. In 2015 in India, around 524 million (39%) practiced open defecation. However, under the SBM mission, due to increase in ‘households with toilets’ only 19 million (1.4%) practiced open defecation in January 2019 [5, 27]. There has also been a 71.58% increase in ‘households with toilets’ from October 2014 to October 2019 in rural areas of the Telangana state in India [27]. The tribal community initially obstructed the effective delivery of health care services; however, after involvement of the local stakeholders and tribal leaders, the acceptance towards medical treatment and community health services improved. Notwithstanding, most of the patients were still reluctant and did not consent for giving stool specimens for laboratory diagnosis. Establishing a rapport with the reticent tribal community was a major challenge faced by the outbreak investigation team. Lack of microbiological aetiology confirmation of the outbreak remained a limitation of the investigation due to limited stool samples and laboratory-capacity constraints of the remote area. Recognizing the pivotal importance of SDGs, national health policy of India (2017) has set the health-related cross-sectorial goal “access to safe water and sanitation to all by 2020†[28]. Greater political and financial commitment towards resource-limited remote tribal areas with effective community mobilization is required to accelerate the public health interventions to improve WASH and to prevent ADD outbreaks in the future. Conclusion This was a community-wide acute diarrheal disease outbreak with high village attack rate in a remote tribal village of Telangana with poor availability of safe water, sanitation and hygiene (WASH) facilities. A rapid and systematic epidemiological investigation identified drinking of faecal-contaminated water from the shallow borewells as the leading cause for this outbreak. These borewells were likely contaminated from runoff after rain from open defecation areas located on a downhill slope. Prompt and targeted public health action contained the number of cases.

What is the date of this event?

85

A community-wide acute diarrheal disease outbreak associated with drinking contaminated water from shallow bore-wells in a tribal village

Abstract Background: In 2016, India reported 709 acute diarrheal disease (ADD) outbreaks (> 25% of all outbreaks). Tribal populations are at higher risk with 27% not having accessibility to safe drinking water and 75% households not having toilets. On June 26, 2017 Pedda-Gujjul-Thanda, a tribal village reported an acute diarrheal disease (ADD) outbreak. We investigated to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods: We defined a case as ≥3 loose stools within 24 h in Pedda-Gujjul-Thanda residents from June 24–30, 2017. We identified cases by reviewing hospital records and house-to-house survey. We conducted a retrospective cohort study and collected stool samples for culture. We assessed drinking water supply and sanitation practices and tested water samples for faecal-contamination. Results: We identified 191 cases (65% females) with median age 36 years (range 4–80 years) and no deaths. The attack-rate (AR) was 37% (191/512). Downhill colonies (located on slope of hilly terrains of the village) reported higher ARs (56%[136/243], p < 0.001) than others (20%[55/269]). Symptoms included diarrhea (100%), fever (17%), vomiting (16%) and abdominal pain (13%). Drinking water from five shallow bore-wells located in downhill colonies was significantly associated with illness (RR = 4.6, 95%CI = 3.4–6.1 and population attributable fraction 61%). In multivariate analysis, drinking water from the shallow bore-wells located in downhill colonies (aOR = 7.9, [95% CI =4.7– 13.2]), illiteracy (aOR =6, [95% CI = 3.6–10.1]), good hand-washing practice (aOR = 0.4, [95%CI = 0.2–0.7]) and household water treatment (aOR = 0.3, [95%CI = 0.2–0.5]) were significantly associated with illness. Two stool cultures were negative for Vibrio cholerae. Heavy rainfall was reported from June 22–24. Five of six water samples collected from shallow bore-wells located in downhill colonies were positive for faecal contamination. Conclusion: An ADD outbreak with high attack rate in a remote tribal village was associated with drinking water from shallow downhill bore-wells, likely contaminated via runoff from open defecation areas after heavy rains. Based on our recommendations, immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families, and as long-term public health measures construction of house-hold latrines and piped-water supply initiated. Keywords: Acute diarrheal disease, Outbreak, Bore-well, Tribal Introduction Globally there are an estimated 1.7 billion cases and 2.2 million deaths from acute diarrheal disease (ADD) every year [1]. In India, the burden is particularly high with more than 13.9 million cases reported in 2016 and 709 ADD outbreaks reported accounting to more than 25% of all outbreaks [2, 3]. Lack of access to safe drinking water and basic sanitation are the leading causes of ADD burden globally and in India. It is estimated that globally 58% of ADD deaths are attributed to inadequate drinking water, sanitation and hygiene [4]. The WHO/UNICEF Joint Monitoring Program for Water Supply, Sanitation and Hygiene (JMP) 2017 report revealed that 844 million people worldwide lack access to basic drinking-water service and 2.3 billion lack basic sanitation services, while 892 million still practiced open defecation [5]. The National Family Health Survey (NFHS-4, 2015–16) reported that in India only 52% of urban households and 18% of rural households have piped water supply, and the main source of water supply among rural households is bore-wells or tube-wells (51%). It has been estimated that 39% of households in India (54% among rural households) have no toilet facility and practicing open defecation [6]. The “indigenous†populations are socially, culturally and economically isolated and usually lack access to basic drinking-water and sanitation services. Therefore, they are vulnerable to ADD outbreaks and other emerging and re-emerging diseases [7]. The United Nations estimates that there are 370 million indigenous people existing across 90 countries of the world. They constitute 5% of the world population but 15% of the poorest [8]. India alone houses more than 705 such indigenous groups termed as Scheduled Tribes. As per the Census 2011, the total Scheduled Tribe population of India is 10.43 crore with a significant proportion of them living in rural areas [9]. On June 26, 2017, Kama-reddy district of Telangana state reported 55 ADD cases from the Pedda-GujjulThanda village. We conducted the outbreak investigation to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods Setting Pedda-Gujjul-Thanda village is a small tribal village with a total population of 563. The village is remotely located as an isolated community with a hilly terrain and is resource-limited with poor accessibility to sanitation and hygiene facilities. The nearest health care facility available for the residents is located at a distance of 10 km from the village. Case definition We defined a case as three or more loose stools within 24 h in a resident of the Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Case finding To find cases, we reviewed medical records of local health care facilities accessed by village residents in the nearby town. We conducted a medical camp in the village during the outbreak period for five days. We conducted a house-to-house survey in the village to find more cases, which are niether reported to health facility nor medical camp. Retrospective cohort study We conducted a retrospective cohort study to identify risk factors associated with illness. We defined the cohort as residents of Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Village resident was the unit of analysis. For data collection, we trained five teams of local paramedical staff. Using a pre-structured questionnaire, we collected data on demographic characteristics and risk factors related to drinking water, sanitation and hygiene. Good hand-washing practice was defined as reported washing of hands with soap and water every time after defecation and before eating. A bore-well less than 30-m-deep, as assessed from the records of village administration, was considered a shallow bore-well. Laboratory and environmental investigations Two stool samples were collected by the treating physician from admitted patients on the first day of hospital admission and transported to the state reference laboratory within two hours in Cary-Blair transport medium. The samples were cultured for Vibrio cholerae, Salmonella and Shigella on nutrient agar, MacConkey agar and deoxycholate citrate agar. Enteric pathogens were identified by biochemical reaction and by agglutination with anti-sera. We collected details of recent rainfall and conducted an environmental survey with household as sampling unit to assess drinking water, sanitation and hygiene practices. We assessed availability of residual chlorine in all village bore-wells and tested four of five bore-wells in the most affected colonies for faecal contamination by H2S method in field. Water was filled up to the ‘fill line’ of the sample bottle and incubated at room temperature (250–370 C) for 36–48 h and observed for colour change in the medium. A water sample was suspected to be contaminated with faecal matter, if it turned black [10, 11]. Because of limited supplies, we were unable to assess the fifth bore-well. Data analysis We analysed the data to describe the occurrence of cases over time, place, and person. We calculated relative risks (RR) with 95% confidence intervals (CI), population attributable risk percentages and conducted multiple logistic regression analysis with the dependent variables including consumption of shallow-downhill bore-well water, report of visible contaminants like mud in drinking water, illiteracy, household water treatment and good hand-washing practice. We used Epi Info version 7.2 for statistical analysis. Results Descriptive epidemiology We identified 191 ADD cases (65% females), with a village attack rate (AR) of 37% (191/512). The attack rate increased with age, with highest among > 60-year age group (55%) and lowest among children under-10 years (11%) (Table 1). No deaths were reported. In addition to diarrhea, cases presented with fever (17%), vomiting (16%) and abdominal pain (13%). 72% (138/191) cases reported to health care facilities and the medical camp conducted in the village. Among the 191 cases, 159 (83%) had mild illness treated with oral rehydration solution; 30 (16%) had moderate dehydration treated with intravenous fluids on out-patient basis, and 2 (1%) with severe dehydration were admitted in the district hospital for treatment with antibiotics (metronidazole and ciprofloxacin) and intravenous fluids. Cases started reported on June 26, 2017, with onset of symptoms from 24 June 2017. Maximum cases were reported on June 27, 2017, and no new cases were reported after June 30, 2017 (Fig. 1). The tribal population in the village had four sub-tribes namely Katroth, Badhawath, Nenawath, Baromath who resided in seven geographically demarcated colonies (labelled as A to G). Katroth sub-tribe resided in colonies A, B and G; Badhawath in colonies C and D; Nenawath in colony E and Baromath in colony F (Table 2). Colonies B and C had higher attack rates (65 and 47% respectively) as compared to other colonies (Fig. 2). Retrospective cohort study Among 563 village residents, 512 (91%) participated in the study. Among the 512 participants, median age was 28 years (range 1–80 years) with 52% females; 50% reported as illiterate with agriculture as the main source of livelihood for 76%. We analysed possible risk factors associated with ADD (Table 3). Drinking water from bore-well groundwater (vs canned water) was found significantly associated with ADD (RR = 12.7; 95% CI = 1.8–87.4). However, only 32 (6%) residents in the village used canned water and bore-well groundwater was the predominant source of water supply. Therefore, we analysed the water sources further, by location and type of bore-wells. Residents who used any of the five shallow bore-wells located downhill were significantly at higher risk (RR = 4.6; 95% CI = 3.4–6.1) and deep bore-wells were protective (RR = 0.4; 95% CI = 0.2–0.9). Report of visible contaminants like mud in drinking water (aOR = 4; 95% CI = 2.1–7.6) and illiteracy (aOR = 3.6; 95% CI = 3.5–10.1) were significantly associated with illness; and household water treatment (done either by boiling or use of candle filters) (aOR = 0.4; 95% CI = 0.2–0.7) and good hand-washing practice (aOR = 0.2; 95% CI = 0.1–0.5) were found protective. Laboratory and environmental results Stool samples collected from two hospitalized cases showed no growth for Vibrio cholerae, Salmonella and Shigella on culture. Among 110 households, 100 (91%) were available for environmental survey. Among the 100 houses surveyed, 79 (79%) were kutcha (low quality) type, made of mud, thatch and other low-quality material. Only 5 (5%) households had a designated toilet at home while the remaining 95 (95%) practiced open defecation at a site located on the slope of the hill behind the downhill colonies B and C (Figs. 2 and 3). Bore-wells were the main source of drinking water supply for 93 (93%) households. There were two deep borewells provided by the village administration and 17 shallow type bore-wells privately constructed by village residents. Five of these 17 (30%) shallow bore-wells were located in colonies B and C, on the downhill slope below the open defecation site. Plastic pipelines from the shallow wells were improperly installed with leakages at multiple points. There was no facility at source, for chlorination or any other mode of purification. Thirty households (30%) treated the water before consumption either by boiling or by use of candle filters. There was no routine drinking water surveillance in place by any authority for assessing the quality and fitness for drinking water. There was no residual chlorine found in any water samples. Three of four drinking water samples from bore-wells of most affected colonies (B and C) indicated faecal contamination by H2S field testing. There was heavy rainfall (average 65 mm in a day) from 22 to 24 June 2017. Prevention and control measures undertaken to contain the outbreak The village residents were discouraged from using shallow bore-well water and were provided with safe canned drinking water until all leakages were secured. Leakages in water supply from the bore-wells were identified and secured. Chlorine tablets were distributed for household level water disinfection. We informed the residents to avoid open defecation near drinking water sources and residential premises. Public health staff conducted health education daily to improve awareness among the villagers about water, sanitation, and hygiene. After active implementation of these control measures, cases declined rapidly in the village (Fig. 1). Discussion A rapid systematic epidemiological investigation of this outbreak identified water contamination points and likely mode of contamination. Based on these findings and our recommendations, the local health department instituted immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families. Effective implementation of public health measures limited the exposure of the community to contaminated water source resulting in rapid containment of the outbreak. Waterborne disease outbreaks tend to have cases spread over a time-period due to ongoing exposure to the contaminated water. In contrast, the pattern of epidemic curve in the present waterborne disease outbreak resembled that of food-borne with a point source exposure. Heavy rains contributed to the run-off of water from the open defecation site into the ground water of shallow wells located on slopes of hilly terrain resulting in heavy contamination and sudden rise of cases. Rapid control measures in the small village, implemented effectively within a short period of time, may have led to rapid decline of cases. The available epidemiological evidence also did not support generation of hypothesis of food-borne origin of the outbreak. In an outbreak reported among school children in Northern Greece in 2012, investigation revealed a waterborne viral gastroenteritis outbreak with a point source pattern, due to consumption of heavily contaminated water from a tap, which was not in use for two weeks during Christmas vacation [12]. Attack rate was high in this outbreak (37%), possibly due to exposure to high pathogen load subsequent to gross faecal contamination of water sources. In the absence of other alternative water sources, this tribal community was exclusively dependent on the contaminated water source for drinking, therefore exposing a large section of the community to risk. Geetha et al. analysed 32 diarrheal outbreaks in south India in non- tribal communities and reported lower attack rates varying from 0.6 to 21.5% [13]. However, tribal populations in India such as in Pedda-Gujjul-Thanda are marginalized with poor availability of WASH facilities [14]. This vulnerable tribal population continues to be at higher risk for ADD outbreaks with 27% not having access to safe drinking water and 75% of households not having toilets [15]. They need special assistance schemes from the government to enable them overcome poor accessibility to WASH facilities and secure healthy living [16]. Due to inadequate availability of communally managed safe public water points by the local authority, this community in Pedda-Gujjul-Thanda village was dependent on privately constructed shallow bore-wells for water supply. These are economical but likely to be unsafe. In this outbreak, open defecation site was present on the downhill slope in proximity to the residential premises and water resources, increasing the risk of drinking water contamination. Among the entire village population, 61% of ADD cases were attributable to drinking water from the ‘shallow downhill bore-wells’ (Population Attributable Fraction 61%), which was also evident from rapid outbreak containment following the elimination of exposure to this single risk factor. Since this exposure factor is amenable to long-term public health intervention, permanent elimination of shallow downhill borewells as water source was recommended, replacing them with properly secured deep bore-wells. Shallow bore-wells are known for their susceptibility to contamination from surface land-use activities [17, 18]. Studies have found levels of E. coli and enteric viruses to be high in shallow sources of ground water especially when they are in close proximity to polluting sources [19–21]. Consumption of ground water from shallow bore-wells with no purification facility increases the risk of diarrhea outbreaks manifold [22, 23]. A metaanalytic study of water-borne diarrheal disease outbreaks in China reported that 78 of 85 (92%) outbreaks (between year 1987 to 2014) were due to poor sanitary conditions of wells with lavatories/septic tanks nearby and lack of purification facilities [24]. In developed countries and urban areas of developing countries, as water supply and sanitation have improved dramatically over a period of time, such outbreaks were rarely reported in the recent past. The largest E. coli O157 outbreak in United States occurred in 1999 at a county fair (781 ill persons and 2 deaths) was due to groundwater source from a temporary unregulated well at the fairground [25]. Our findings have implication for India’s progress towards United Nation’s Sustainable Development Goal (SDG) 6 and India’s nation-wide campaign ‘Swachh Bharat Mission (SBM)’ to ensure availability and management of water and sanitation for all. SDG 6 aims at achieving universal access to basic sanitation service by 2030; and it has been reported that between 2000 and 2017, the proportion lacking even a basic sanitation service decreased from 44 to 27% [26]. SBM aims to achieve an open-defecation free status in rural areas through the construction of household-owned and community-owned toilets and establishing an accountable mechanism of monitoring toilet use. In 2015 in India, around 524 million (39%) practiced open defecation. However, under the SBM mission, due to increase in ‘households with toilets’ only 19 million (1.4%) practiced open defecation in January 2019 [5, 27]. There has also been a 71.58% increase in ‘households with toilets’ from October 2014 to October 2019 in rural areas of the Telangana state in India [27]. The tribal community initially obstructed the effective delivery of health care services; however, after involvement of the local stakeholders and tribal leaders, the acceptance towards medical treatment and community health services improved. Notwithstanding, most of the patients were still reluctant and did not consent for giving stool specimens for laboratory diagnosis. Establishing a rapport with the reticent tribal community was a major challenge faced by the outbreak investigation team. Lack of microbiological aetiology confirmation of the outbreak remained a limitation of the investigation due to limited stool samples and laboratory-capacity constraints of the remote area. Recognizing the pivotal importance of SDGs, national health policy of India (2017) has set the health-related cross-sectorial goal “access to safe water and sanitation to all by 2020†[28]. Greater political and financial commitment towards resource-limited remote tribal areas with effective community mobilization is required to accelerate the public health interventions to improve WASH and to prevent ADD outbreaks in the future. Conclusion This was a community-wide acute diarrheal disease outbreak with high village attack rate in a remote tribal village of Telangana with poor availability of safe water, sanitation and hygiene (WASH) facilities. A rapid and systematic epidemiological investigation identified drinking of faecal-contaminated water from the shallow borewells as the leading cause for this outbreak. These borewells were likely contaminated from runoff after rain from open defecation areas located on a downhill slope. Prompt and targeted public health action contained the number of cases.

How long was the event?

86

A community-wide acute diarrheal disease outbreak associated with drinking contaminated water from shallow bore-wells in a tribal village

Abstract Background: In 2016, India reported 709 acute diarrheal disease (ADD) outbreaks (> 25% of all outbreaks). Tribal populations are at higher risk with 27% not having accessibility to safe drinking water and 75% households not having toilets. On June 26, 2017 Pedda-Gujjul-Thanda, a tribal village reported an acute diarrheal disease (ADD) outbreak. We investigated to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods: We defined a case as ≥3 loose stools within 24 h in Pedda-Gujjul-Thanda residents from June 24–30, 2017. We identified cases by reviewing hospital records and house-to-house survey. We conducted a retrospective cohort study and collected stool samples for culture. We assessed drinking water supply and sanitation practices and tested water samples for faecal-contamination. Results: We identified 191 cases (65% females) with median age 36 years (range 4–80 years) and no deaths. The attack-rate (AR) was 37% (191/512). Downhill colonies (located on slope of hilly terrains of the village) reported higher ARs (56%[136/243], p < 0.001) than others (20%[55/269]). Symptoms included diarrhea (100%), fever (17%), vomiting (16%) and abdominal pain (13%). Drinking water from five shallow bore-wells located in downhill colonies was significantly associated with illness (RR = 4.6, 95%CI = 3.4–6.1 and population attributable fraction 61%). In multivariate analysis, drinking water from the shallow bore-wells located in downhill colonies (aOR = 7.9, [95% CI =4.7– 13.2]), illiteracy (aOR =6, [95% CI = 3.6–10.1]), good hand-washing practice (aOR = 0.4, [95%CI = 0.2–0.7]) and household water treatment (aOR = 0.3, [95%CI = 0.2–0.5]) were significantly associated with illness. Two stool cultures were negative for Vibrio cholerae. Heavy rainfall was reported from June 22–24. Five of six water samples collected from shallow bore-wells located in downhill colonies were positive for faecal contamination. Conclusion: An ADD outbreak with high attack rate in a remote tribal village was associated with drinking water from shallow downhill bore-wells, likely contaminated via runoff from open defecation areas after heavy rains. Based on our recommendations, immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families, and as long-term public health measures construction of house-hold latrines and piped-water supply initiated. Keywords: Acute diarrheal disease, Outbreak, Bore-well, Tribal Introduction Globally there are an estimated 1.7 billion cases and 2.2 million deaths from acute diarrheal disease (ADD) every year [1]. In India, the burden is particularly high with more than 13.9 million cases reported in 2016 and 709 ADD outbreaks reported accounting to more than 25% of all outbreaks [2, 3]. Lack of access to safe drinking water and basic sanitation are the leading causes of ADD burden globally and in India. It is estimated that globally 58% of ADD deaths are attributed to inadequate drinking water, sanitation and hygiene [4]. The WHO/UNICEF Joint Monitoring Program for Water Supply, Sanitation and Hygiene (JMP) 2017 report revealed that 844 million people worldwide lack access to basic drinking-water service and 2.3 billion lack basic sanitation services, while 892 million still practiced open defecation [5]. The National Family Health Survey (NFHS-4, 2015–16) reported that in India only 52% of urban households and 18% of rural households have piped water supply, and the main source of water supply among rural households is bore-wells or tube-wells (51%). It has been estimated that 39% of households in India (54% among rural households) have no toilet facility and practicing open defecation [6]. The “indigenous†populations are socially, culturally and economically isolated and usually lack access to basic drinking-water and sanitation services. Therefore, they are vulnerable to ADD outbreaks and other emerging and re-emerging diseases [7]. The United Nations estimates that there are 370 million indigenous people existing across 90 countries of the world. They constitute 5% of the world population but 15% of the poorest [8]. India alone houses more than 705 such indigenous groups termed as Scheduled Tribes. As per the Census 2011, the total Scheduled Tribe population of India is 10.43 crore with a significant proportion of them living in rural areas [9]. On June 26, 2017, Kama-reddy district of Telangana state reported 55 ADD cases from the Pedda-GujjulThanda village. We conducted the outbreak investigation to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods Setting Pedda-Gujjul-Thanda village is a small tribal village with a total population of 563. The village is remotely located as an isolated community with a hilly terrain and is resource-limited with poor accessibility to sanitation and hygiene facilities. The nearest health care facility available for the residents is located at a distance of 10 km from the village. Case definition We defined a case as three or more loose stools within 24 h in a resident of the Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Case finding To find cases, we reviewed medical records of local health care facilities accessed by village residents in the nearby town. We conducted a medical camp in the village during the outbreak period for five days. We conducted a house-to-house survey in the village to find more cases, which are niether reported to health facility nor medical camp. Retrospective cohort study We conducted a retrospective cohort study to identify risk factors associated with illness. We defined the cohort as residents of Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Village resident was the unit of analysis. For data collection, we trained five teams of local paramedical staff. Using a pre-structured questionnaire, we collected data on demographic characteristics and risk factors related to drinking water, sanitation and hygiene. Good hand-washing practice was defined as reported washing of hands with soap and water every time after defecation and before eating. A bore-well less than 30-m-deep, as assessed from the records of village administration, was considered a shallow bore-well. Laboratory and environmental investigations Two stool samples were collected by the treating physician from admitted patients on the first day of hospital admission and transported to the state reference laboratory within two hours in Cary-Blair transport medium. The samples were cultured for Vibrio cholerae, Salmonella and Shigella on nutrient agar, MacConkey agar and deoxycholate citrate agar. Enteric pathogens were identified by biochemical reaction and by agglutination with anti-sera. We collected details of recent rainfall and conducted an environmental survey with household as sampling unit to assess drinking water, sanitation and hygiene practices. We assessed availability of residual chlorine in all village bore-wells and tested four of five bore-wells in the most affected colonies for faecal contamination by H2S method in field. Water was filled up to the ‘fill line’ of the sample bottle and incubated at room temperature (250–370 C) for 36–48 h and observed for colour change in the medium. A water sample was suspected to be contaminated with faecal matter, if it turned black [10, 11]. Because of limited supplies, we were unable to assess the fifth bore-well. Data analysis We analysed the data to describe the occurrence of cases over time, place, and person. We calculated relative risks (RR) with 95% confidence intervals (CI), population attributable risk percentages and conducted multiple logistic regression analysis with the dependent variables including consumption of shallow-downhill bore-well water, report of visible contaminants like mud in drinking water, illiteracy, household water treatment and good hand-washing practice. We used Epi Info version 7.2 for statistical analysis. Results Descriptive epidemiology We identified 191 ADD cases (65% females), with a village attack rate (AR) of 37% (191/512). The attack rate increased with age, with highest among > 60-year age group (55%) and lowest among children under-10 years (11%) (Table 1). No deaths were reported. In addition to diarrhea, cases presented with fever (17%), vomiting (16%) and abdominal pain (13%). 72% (138/191) cases reported to health care facilities and the medical camp conducted in the village. Among the 191 cases, 159 (83%) had mild illness treated with oral rehydration solution; 30 (16%) had moderate dehydration treated with intravenous fluids on out-patient basis, and 2 (1%) with severe dehydration were admitted in the district hospital for treatment with antibiotics (metronidazole and ciprofloxacin) and intravenous fluids. Cases started reported on June 26, 2017, with onset of symptoms from 24 June 2017. Maximum cases were reported on June 27, 2017, and no new cases were reported after June 30, 2017 (Fig. 1). The tribal population in the village had four sub-tribes namely Katroth, Badhawath, Nenawath, Baromath who resided in seven geographically demarcated colonies (labelled as A to G). Katroth sub-tribe resided in colonies A, B and G; Badhawath in colonies C and D; Nenawath in colony E and Baromath in colony F (Table 2). Colonies B and C had higher attack rates (65 and 47% respectively) as compared to other colonies (Fig. 2). Retrospective cohort study Among 563 village residents, 512 (91%) participated in the study. Among the 512 participants, median age was 28 years (range 1–80 years) with 52% females; 50% reported as illiterate with agriculture as the main source of livelihood for 76%. We analysed possible risk factors associated with ADD (Table 3). Drinking water from bore-well groundwater (vs canned water) was found significantly associated with ADD (RR = 12.7; 95% CI = 1.8–87.4). However, only 32 (6%) residents in the village used canned water and bore-well groundwater was the predominant source of water supply. Therefore, we analysed the water sources further, by location and type of bore-wells. Residents who used any of the five shallow bore-wells located downhill were significantly at higher risk (RR = 4.6; 95% CI = 3.4–6.1) and deep bore-wells were protective (RR = 0.4; 95% CI = 0.2–0.9). Report of visible contaminants like mud in drinking water (aOR = 4; 95% CI = 2.1–7.6) and illiteracy (aOR = 3.6; 95% CI = 3.5–10.1) were significantly associated with illness; and household water treatment (done either by boiling or use of candle filters) (aOR = 0.4; 95% CI = 0.2–0.7) and good hand-washing practice (aOR = 0.2; 95% CI = 0.1–0.5) were found protective. Laboratory and environmental results Stool samples collected from two hospitalized cases showed no growth for Vibrio cholerae, Salmonella and Shigella on culture. Among 110 households, 100 (91%) were available for environmental survey. Among the 100 houses surveyed, 79 (79%) were kutcha (low quality) type, made of mud, thatch and other low-quality material. Only 5 (5%) households had a designated toilet at home while the remaining 95 (95%) practiced open defecation at a site located on the slope of the hill behind the downhill colonies B and C (Figs. 2 and 3). Bore-wells were the main source of drinking water supply for 93 (93%) households. There were two deep borewells provided by the village administration and 17 shallow type bore-wells privately constructed by village residents. Five of these 17 (30%) shallow bore-wells were located in colonies B and C, on the downhill slope below the open defecation site. Plastic pipelines from the shallow wells were improperly installed with leakages at multiple points. There was no facility at source, for chlorination or any other mode of purification. Thirty households (30%) treated the water before consumption either by boiling or by use of candle filters. There was no routine drinking water surveillance in place by any authority for assessing the quality and fitness for drinking water. There was no residual chlorine found in any water samples. Three of four drinking water samples from bore-wells of most affected colonies (B and C) indicated faecal contamination by H2S field testing. There was heavy rainfall (average 65 mm in a day) from 22 to 24 June 2017. Prevention and control measures undertaken to contain the outbreak The village residents were discouraged from using shallow bore-well water and were provided with safe canned drinking water until all leakages were secured. Leakages in water supply from the bore-wells were identified and secured. Chlorine tablets were distributed for household level water disinfection. We informed the residents to avoid open defecation near drinking water sources and residential premises. Public health staff conducted health education daily to improve awareness among the villagers about water, sanitation, and hygiene. After active implementation of these control measures, cases declined rapidly in the village (Fig. 1). Discussion A rapid systematic epidemiological investigation of this outbreak identified water contamination points and likely mode of contamination. Based on these findings and our recommendations, the local health department instituted immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families. Effective implementation of public health measures limited the exposure of the community to contaminated water source resulting in rapid containment of the outbreak. Waterborne disease outbreaks tend to have cases spread over a time-period due to ongoing exposure to the contaminated water. In contrast, the pattern of epidemic curve in the present waterborne disease outbreak resembled that of food-borne with a point source exposure. Heavy rains contributed to the run-off of water from the open defecation site into the ground water of shallow wells located on slopes of hilly terrain resulting in heavy contamination and sudden rise of cases. Rapid control measures in the small village, implemented effectively within a short period of time, may have led to rapid decline of cases. The available epidemiological evidence also did not support generation of hypothesis of food-borne origin of the outbreak. In an outbreak reported among school children in Northern Greece in 2012, investigation revealed a waterborne viral gastroenteritis outbreak with a point source pattern, due to consumption of heavily contaminated water from a tap, which was not in use for two weeks during Christmas vacation [12]. Attack rate was high in this outbreak (37%), possibly due to exposure to high pathogen load subsequent to gross faecal contamination of water sources. In the absence of other alternative water sources, this tribal community was exclusively dependent on the contaminated water source for drinking, therefore exposing a large section of the community to risk. Geetha et al. analysed 32 diarrheal outbreaks in south India in non- tribal communities and reported lower attack rates varying from 0.6 to 21.5% [13]. However, tribal populations in India such as in Pedda-Gujjul-Thanda are marginalized with poor availability of WASH facilities [14]. This vulnerable tribal population continues to be at higher risk for ADD outbreaks with 27% not having access to safe drinking water and 75% of households not having toilets [15]. They need special assistance schemes from the government to enable them overcome poor accessibility to WASH facilities and secure healthy living [16]. Due to inadequate availability of communally managed safe public water points by the local authority, this community in Pedda-Gujjul-Thanda village was dependent on privately constructed shallow bore-wells for water supply. These are economical but likely to be unsafe. In this outbreak, open defecation site was present on the downhill slope in proximity to the residential premises and water resources, increasing the risk of drinking water contamination. Among the entire village population, 61% of ADD cases were attributable to drinking water from the ‘shallow downhill bore-wells’ (Population Attributable Fraction 61%), which was also evident from rapid outbreak containment following the elimination of exposure to this single risk factor. Since this exposure factor is amenable to long-term public health intervention, permanent elimination of shallow downhill borewells as water source was recommended, replacing them with properly secured deep bore-wells. Shallow bore-wells are known for their susceptibility to contamination from surface land-use activities [17, 18]. Studies have found levels of E. coli and enteric viruses to be high in shallow sources of ground water especially when they are in close proximity to polluting sources [19–21]. Consumption of ground water from shallow bore-wells with no purification facility increases the risk of diarrhea outbreaks manifold [22, 23]. A metaanalytic study of water-borne diarrheal disease outbreaks in China reported that 78 of 85 (92%) outbreaks (between year 1987 to 2014) were due to poor sanitary conditions of wells with lavatories/septic tanks nearby and lack of purification facilities [24]. In developed countries and urban areas of developing countries, as water supply and sanitation have improved dramatically over a period of time, such outbreaks were rarely reported in the recent past. The largest E. coli O157 outbreak in United States occurred in 1999 at a county fair (781 ill persons and 2 deaths) was due to groundwater source from a temporary unregulated well at the fairground [25]. Our findings have implication for India’s progress towards United Nation’s Sustainable Development Goal (SDG) 6 and India’s nation-wide campaign ‘Swachh Bharat Mission (SBM)’ to ensure availability and management of water and sanitation for all. SDG 6 aims at achieving universal access to basic sanitation service by 2030; and it has been reported that between 2000 and 2017, the proportion lacking even a basic sanitation service decreased from 44 to 27% [26]. SBM aims to achieve an open-defecation free status in rural areas through the construction of household-owned and community-owned toilets and establishing an accountable mechanism of monitoring toilet use. In 2015 in India, around 524 million (39%) practiced open defecation. However, under the SBM mission, due to increase in ‘households with toilets’ only 19 million (1.4%) practiced open defecation in January 2019 [5, 27]. There has also been a 71.58% increase in ‘households with toilets’ from October 2014 to October 2019 in rural areas of the Telangana state in India [27]. The tribal community initially obstructed the effective delivery of health care services; however, after involvement of the local stakeholders and tribal leaders, the acceptance towards medical treatment and community health services improved. Notwithstanding, most of the patients were still reluctant and did not consent for giving stool specimens for laboratory diagnosis. Establishing a rapport with the reticent tribal community was a major challenge faced by the outbreak investigation team. Lack of microbiological aetiology confirmation of the outbreak remained a limitation of the investigation due to limited stool samples and laboratory-capacity constraints of the remote area. Recognizing the pivotal importance of SDGs, national health policy of India (2017) has set the health-related cross-sectorial goal “access to safe water and sanitation to all by 2020†[28]. Greater political and financial commitment towards resource-limited remote tribal areas with effective community mobilization is required to accelerate the public health interventions to improve WASH and to prevent ADD outbreaks in the future. Conclusion This was a community-wide acute diarrheal disease outbreak with high village attack rate in a remote tribal village of Telangana with poor availability of safe water, sanitation and hygiene (WASH) facilities. A rapid and systematic epidemiological investigation identified drinking of faecal-contaminated water from the shallow borewells as the leading cause for this outbreak. These borewells were likely contaminated from runoff after rain from open defecation areas located on a downhill slope. Prompt and targeted public health action contained the number of cases.

How long did the event last?

87

A community-wide acute diarrheal disease outbreak associated with drinking contaminated water from shallow bore-wells in a tribal village

Abstract Background: In 2016, India reported 709 acute diarrheal disease (ADD) outbreaks (> 25% of all outbreaks). Tribal populations are at higher risk with 27% not having accessibility to safe drinking water and 75% households not having toilets. On June 26, 2017 Pedda-Gujjul-Thanda, a tribal village reported an acute diarrheal disease (ADD) outbreak. We investigated to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods: We defined a case as ≥3 loose stools within 24 h in Pedda-Gujjul-Thanda residents from June 24–30, 2017. We identified cases by reviewing hospital records and house-to-house survey. We conducted a retrospective cohort study and collected stool samples for culture. We assessed drinking water supply and sanitation practices and tested water samples for faecal-contamination. Results: We identified 191 cases (65% females) with median age 36 years (range 4–80 years) and no deaths. The attack-rate (AR) was 37% (191/512). Downhill colonies (located on slope of hilly terrains of the village) reported higher ARs (56%[136/243], p < 0.001) than others (20%[55/269]). Symptoms included diarrhea (100%), fever (17%), vomiting (16%) and abdominal pain (13%). Drinking water from five shallow bore-wells located in downhill colonies was significantly associated with illness (RR = 4.6, 95%CI = 3.4–6.1 and population attributable fraction 61%). In multivariate analysis, drinking water from the shallow bore-wells located in downhill colonies (aOR = 7.9, [95% CI =4.7– 13.2]), illiteracy (aOR =6, [95% CI = 3.6–10.1]), good hand-washing practice (aOR = 0.4, [95%CI = 0.2–0.7]) and household water treatment (aOR = 0.3, [95%CI = 0.2–0.5]) were significantly associated with illness. Two stool cultures were negative for Vibrio cholerae. Heavy rainfall was reported from June 22–24. Five of six water samples collected from shallow bore-wells located in downhill colonies were positive for faecal contamination. Conclusion: An ADD outbreak with high attack rate in a remote tribal village was associated with drinking water from shallow downhill bore-wells, likely contaminated via runoff from open defecation areas after heavy rains. Based on our recommendations, immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families, and as long-term public health measures construction of house-hold latrines and piped-water supply initiated. Keywords: Acute diarrheal disease, Outbreak, Bore-well, Tribal Introduction Globally there are an estimated 1.7 billion cases and 2.2 million deaths from acute diarrheal disease (ADD) every year [1]. In India, the burden is particularly high with more than 13.9 million cases reported in 2016 and 709 ADD outbreaks reported accounting to more than 25% of all outbreaks [2, 3]. Lack of access to safe drinking water and basic sanitation are the leading causes of ADD burden globally and in India. It is estimated that globally 58% of ADD deaths are attributed to inadequate drinking water, sanitation and hygiene [4]. The WHO/UNICEF Joint Monitoring Program for Water Supply, Sanitation and Hygiene (JMP) 2017 report revealed that 844 million people worldwide lack access to basic drinking-water service and 2.3 billion lack basic sanitation services, while 892 million still practiced open defecation [5]. The National Family Health Survey (NFHS-4, 2015–16) reported that in India only 52% of urban households and 18% of rural households have piped water supply, and the main source of water supply among rural households is bore-wells or tube-wells (51%). It has been estimated that 39% of households in India (54% among rural households) have no toilet facility and practicing open defecation [6]. The “indigenous†populations are socially, culturally and economically isolated and usually lack access to basic drinking-water and sanitation services. Therefore, they are vulnerable to ADD outbreaks and other emerging and re-emerging diseases [7]. The United Nations estimates that there are 370 million indigenous people existing across 90 countries of the world. They constitute 5% of the world population but 15% of the poorest [8]. India alone houses more than 705 such indigenous groups termed as Scheduled Tribes. As per the Census 2011, the total Scheduled Tribe population of India is 10.43 crore with a significant proportion of them living in rural areas [9]. On June 26, 2017, Kama-reddy district of Telangana state reported 55 ADD cases from the Pedda-GujjulThanda village. We conducted the outbreak investigation to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods Setting Pedda-Gujjul-Thanda village is a small tribal village with a total population of 563. The village is remotely located as an isolated community with a hilly terrain and is resource-limited with poor accessibility to sanitation and hygiene facilities. The nearest health care facility available for the residents is located at a distance of 10 km from the village. Case definition We defined a case as three or more loose stools within 24 h in a resident of the Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Case finding To find cases, we reviewed medical records of local health care facilities accessed by village residents in the nearby town. We conducted a medical camp in the village during the outbreak period for five days. We conducted a house-to-house survey in the village to find more cases, which are niether reported to health facility nor medical camp. Retrospective cohort study We conducted a retrospective cohort study to identify risk factors associated with illness. We defined the cohort as residents of Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Village resident was the unit of analysis. For data collection, we trained five teams of local paramedical staff. Using a pre-structured questionnaire, we collected data on demographic characteristics and risk factors related to drinking water, sanitation and hygiene. Good hand-washing practice was defined as reported washing of hands with soap and water every time after defecation and before eating. A bore-well less than 30-m-deep, as assessed from the records of village administration, was considered a shallow bore-well. Laboratory and environmental investigations Two stool samples were collected by the treating physician from admitted patients on the first day of hospital admission and transported to the state reference laboratory within two hours in Cary-Blair transport medium. The samples were cultured for Vibrio cholerae, Salmonella and Shigella on nutrient agar, MacConkey agar and deoxycholate citrate agar. Enteric pathogens were identified by biochemical reaction and by agglutination with anti-sera. We collected details of recent rainfall and conducted an environmental survey with household as sampling unit to assess drinking water, sanitation and hygiene practices. We assessed availability of residual chlorine in all village bore-wells and tested four of five bore-wells in the most affected colonies for faecal contamination by H2S method in field. Water was filled up to the ‘fill line’ of the sample bottle and incubated at room temperature (250–370 C) for 36–48 h and observed for colour change in the medium. A water sample was suspected to be contaminated with faecal matter, if it turned black [10, 11]. Because of limited supplies, we were unable to assess the fifth bore-well. Data analysis We analysed the data to describe the occurrence of cases over time, place, and person. We calculated relative risks (RR) with 95% confidence intervals (CI), population attributable risk percentages and conducted multiple logistic regression analysis with the dependent variables including consumption of shallow-downhill bore-well water, report of visible contaminants like mud in drinking water, illiteracy, household water treatment and good hand-washing practice. We used Epi Info version 7.2 for statistical analysis. Results Descriptive epidemiology We identified 191 ADD cases (65% females), with a village attack rate (AR) of 37% (191/512). The attack rate increased with age, with highest among > 60-year age group (55%) and lowest among children under-10 years (11%) (Table 1). No deaths were reported. In addition to diarrhea, cases presented with fever (17%), vomiting (16%) and abdominal pain (13%). 72% (138/191) cases reported to health care facilities and the medical camp conducted in the village. Among the 191 cases, 159 (83%) had mild illness treated with oral rehydration solution; 30 (16%) had moderate dehydration treated with intravenous fluids on out-patient basis, and 2 (1%) with severe dehydration were admitted in the district hospital for treatment with antibiotics (metronidazole and ciprofloxacin) and intravenous fluids. Cases started reported on June 26, 2017, with onset of symptoms from 24 June 2017. Maximum cases were reported on June 27, 2017, and no new cases were reported after June 30, 2017 (Fig. 1). The tribal population in the village had four sub-tribes namely Katroth, Badhawath, Nenawath, Baromath who resided in seven geographically demarcated colonies (labelled as A to G). Katroth sub-tribe resided in colonies A, B and G; Badhawath in colonies C and D; Nenawath in colony E and Baromath in colony F (Table 2). Colonies B and C had higher attack rates (65 and 47% respectively) as compared to other colonies (Fig. 2). Retrospective cohort study Among 563 village residents, 512 (91%) participated in the study. Among the 512 participants, median age was 28 years (range 1–80 years) with 52% females; 50% reported as illiterate with agriculture as the main source of livelihood for 76%. We analysed possible risk factors associated with ADD (Table 3). Drinking water from bore-well groundwater (vs canned water) was found significantly associated with ADD (RR = 12.7; 95% CI = 1.8–87.4). However, only 32 (6%) residents in the village used canned water and bore-well groundwater was the predominant source of water supply. Therefore, we analysed the water sources further, by location and type of bore-wells. Residents who used any of the five shallow bore-wells located downhill were significantly at higher risk (RR = 4.6; 95% CI = 3.4–6.1) and deep bore-wells were protective (RR = 0.4; 95% CI = 0.2–0.9). Report of visible contaminants like mud in drinking water (aOR = 4; 95% CI = 2.1–7.6) and illiteracy (aOR = 3.6; 95% CI = 3.5–10.1) were significantly associated with illness; and household water treatment (done either by boiling or use of candle filters) (aOR = 0.4; 95% CI = 0.2–0.7) and good hand-washing practice (aOR = 0.2; 95% CI = 0.1–0.5) were found protective. Laboratory and environmental results Stool samples collected from two hospitalized cases showed no growth for Vibrio cholerae, Salmonella and Shigella on culture. Among 110 households, 100 (91%) were available for environmental survey. Among the 100 houses surveyed, 79 (79%) were kutcha (low quality) type, made of mud, thatch and other low-quality material. Only 5 (5%) households had a designated toilet at home while the remaining 95 (95%) practiced open defecation at a site located on the slope of the hill behind the downhill colonies B and C (Figs. 2 and 3). Bore-wells were the main source of drinking water supply for 93 (93%) households. There were two deep borewells provided by the village administration and 17 shallow type bore-wells privately constructed by village residents. Five of these 17 (30%) shallow bore-wells were located in colonies B and C, on the downhill slope below the open defecation site. Plastic pipelines from the shallow wells were improperly installed with leakages at multiple points. There was no facility at source, for chlorination or any other mode of purification. Thirty households (30%) treated the water before consumption either by boiling or by use of candle filters. There was no routine drinking water surveillance in place by any authority for assessing the quality and fitness for drinking water. There was no residual chlorine found in any water samples. Three of four drinking water samples from bore-wells of most affected colonies (B and C) indicated faecal contamination by H2S field testing. There was heavy rainfall (average 65 mm in a day) from 22 to 24 June 2017. Prevention and control measures undertaken to contain the outbreak The village residents were discouraged from using shallow bore-well water and were provided with safe canned drinking water until all leakages were secured. Leakages in water supply from the bore-wells were identified and secured. Chlorine tablets were distributed for household level water disinfection. We informed the residents to avoid open defecation near drinking water sources and residential premises. Public health staff conducted health education daily to improve awareness among the villagers about water, sanitation, and hygiene. After active implementation of these control measures, cases declined rapidly in the village (Fig. 1). Discussion A rapid systematic epidemiological investigation of this outbreak identified water contamination points and likely mode of contamination. Based on these findings and our recommendations, the local health department instituted immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families. Effective implementation of public health measures limited the exposure of the community to contaminated water source resulting in rapid containment of the outbreak. Waterborne disease outbreaks tend to have cases spread over a time-period due to ongoing exposure to the contaminated water. In contrast, the pattern of epidemic curve in the present waterborne disease outbreak resembled that of food-borne with a point source exposure. Heavy rains contributed to the run-off of water from the open defecation site into the ground water of shallow wells located on slopes of hilly terrain resulting in heavy contamination and sudden rise of cases. Rapid control measures in the small village, implemented effectively within a short period of time, may have led to rapid decline of cases. The available epidemiological evidence also did not support generation of hypothesis of food-borne origin of the outbreak. In an outbreak reported among school children in Northern Greece in 2012, investigation revealed a waterborne viral gastroenteritis outbreak with a point source pattern, due to consumption of heavily contaminated water from a tap, which was not in use for two weeks during Christmas vacation [12]. Attack rate was high in this outbreak (37%), possibly due to exposure to high pathogen load subsequent to gross faecal contamination of water sources. In the absence of other alternative water sources, this tribal community was exclusively dependent on the contaminated water source for drinking, therefore exposing a large section of the community to risk. Geetha et al. analysed 32 diarrheal outbreaks in south India in non- tribal communities and reported lower attack rates varying from 0.6 to 21.5% [13]. However, tribal populations in India such as in Pedda-Gujjul-Thanda are marginalized with poor availability of WASH facilities [14]. This vulnerable tribal population continues to be at higher risk for ADD outbreaks with 27% not having access to safe drinking water and 75% of households not having toilets [15]. They need special assistance schemes from the government to enable them overcome poor accessibility to WASH facilities and secure healthy living [16]. Due to inadequate availability of communally managed safe public water points by the local authority, this community in Pedda-Gujjul-Thanda village was dependent on privately constructed shallow bore-wells for water supply. These are economical but likely to be unsafe. In this outbreak, open defecation site was present on the downhill slope in proximity to the residential premises and water resources, increasing the risk of drinking water contamination. Among the entire village population, 61% of ADD cases were attributable to drinking water from the ‘shallow downhill bore-wells’ (Population Attributable Fraction 61%), which was also evident from rapid outbreak containment following the elimination of exposure to this single risk factor. Since this exposure factor is amenable to long-term public health intervention, permanent elimination of shallow downhill borewells as water source was recommended, replacing them with properly secured deep bore-wells. Shallow bore-wells are known for their susceptibility to contamination from surface land-use activities [17, 18]. Studies have found levels of E. coli and enteric viruses to be high in shallow sources of ground water especially when they are in close proximity to polluting sources [19–21]. Consumption of ground water from shallow bore-wells with no purification facility increases the risk of diarrhea outbreaks manifold [22, 23]. A metaanalytic study of water-borne diarrheal disease outbreaks in China reported that 78 of 85 (92%) outbreaks (between year 1987 to 2014) were due to poor sanitary conditions of wells with lavatories/septic tanks nearby and lack of purification facilities [24]. In developed countries and urban areas of developing countries, as water supply and sanitation have improved dramatically over a period of time, such outbreaks were rarely reported in the recent past. The largest E. coli O157 outbreak in United States occurred in 1999 at a county fair (781 ill persons and 2 deaths) was due to groundwater source from a temporary unregulated well at the fairground [25]. Our findings have implication for India’s progress towards United Nation’s Sustainable Development Goal (SDG) 6 and India’s nation-wide campaign ‘Swachh Bharat Mission (SBM)’ to ensure availability and management of water and sanitation for all. SDG 6 aims at achieving universal access to basic sanitation service by 2030; and it has been reported that between 2000 and 2017, the proportion lacking even a basic sanitation service decreased from 44 to 27% [26]. SBM aims to achieve an open-defecation free status in rural areas through the construction of household-owned and community-owned toilets and establishing an accountable mechanism of monitoring toilet use. In 2015 in India, around 524 million (39%) practiced open defecation. However, under the SBM mission, due to increase in ‘households with toilets’ only 19 million (1.4%) practiced open defecation in January 2019 [5, 27]. There has also been a 71.58% increase in ‘households with toilets’ from October 2014 to October 2019 in rural areas of the Telangana state in India [27]. The tribal community initially obstructed the effective delivery of health care services; however, after involvement of the local stakeholders and tribal leaders, the acceptance towards medical treatment and community health services improved. Notwithstanding, most of the patients were still reluctant and did not consent for giving stool specimens for laboratory diagnosis. Establishing a rapport with the reticent tribal community was a major challenge faced by the outbreak investigation team. Lack of microbiological aetiology confirmation of the outbreak remained a limitation of the investigation due to limited stool samples and laboratory-capacity constraints of the remote area. Recognizing the pivotal importance of SDGs, national health policy of India (2017) has set the health-related cross-sectorial goal “access to safe water and sanitation to all by 2020†[28]. Greater political and financial commitment towards resource-limited remote tribal areas with effective community mobilization is required to accelerate the public health interventions to improve WASH and to prevent ADD outbreaks in the future. Conclusion This was a community-wide acute diarrheal disease outbreak with high village attack rate in a remote tribal village of Telangana with poor availability of safe water, sanitation and hygiene (WASH) facilities. A rapid and systematic epidemiological investigation identified drinking of faecal-contaminated water from the shallow borewells as the leading cause for this outbreak. These borewells were likely contaminated from runoff after rain from open defecation areas located on a downhill slope. Prompt and targeted public health action contained the number of cases.

In which street did this happen?

88

A community-wide acute diarrheal disease outbreak associated with drinking contaminated water from shallow bore-wells in a tribal village

Abstract Background: In 2016, India reported 709 acute diarrheal disease (ADD) outbreaks (> 25% of all outbreaks). Tribal populations are at higher risk with 27% not having accessibility to safe drinking water and 75% households not having toilets. On June 26, 2017 Pedda-Gujjul-Thanda, a tribal village reported an acute diarrheal disease (ADD) outbreak. We investigated to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods: We defined a case as ≥3 loose stools within 24 h in Pedda-Gujjul-Thanda residents from June 24–30, 2017. We identified cases by reviewing hospital records and house-to-house survey. We conducted a retrospective cohort study and collected stool samples for culture. We assessed drinking water supply and sanitation practices and tested water samples for faecal-contamination. Results: We identified 191 cases (65% females) with median age 36 years (range 4–80 years) and no deaths. The attack-rate (AR) was 37% (191/512). Downhill colonies (located on slope of hilly terrains of the village) reported higher ARs (56%[136/243], p < 0.001) than others (20%[55/269]). Symptoms included diarrhea (100%), fever (17%), vomiting (16%) and abdominal pain (13%). Drinking water from five shallow bore-wells located in downhill colonies was significantly associated with illness (RR = 4.6, 95%CI = 3.4–6.1 and population attributable fraction 61%). In multivariate analysis, drinking water from the shallow bore-wells located in downhill colonies (aOR = 7.9, [95% CI =4.7– 13.2]), illiteracy (aOR =6, [95% CI = 3.6–10.1]), good hand-washing practice (aOR = 0.4, [95%CI = 0.2–0.7]) and household water treatment (aOR = 0.3, [95%CI = 0.2–0.5]) were significantly associated with illness. Two stool cultures were negative for Vibrio cholerae. Heavy rainfall was reported from June 22–24. Five of six water samples collected from shallow bore-wells located in downhill colonies were positive for faecal contamination. Conclusion: An ADD outbreak with high attack rate in a remote tribal village was associated with drinking water from shallow downhill bore-wells, likely contaminated via runoff from open defecation areas after heavy rains. Based on our recommendations, immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families, and as long-term public health measures construction of house-hold latrines and piped-water supply initiated. Keywords: Acute diarrheal disease, Outbreak, Bore-well, Tribal Introduction Globally there are an estimated 1.7 billion cases and 2.2 million deaths from acute diarrheal disease (ADD) every year [1]. In India, the burden is particularly high with more than 13.9 million cases reported in 2016 and 709 ADD outbreaks reported accounting to more than 25% of all outbreaks [2, 3]. Lack of access to safe drinking water and basic sanitation are the leading causes of ADD burden globally and in India. It is estimated that globally 58% of ADD deaths are attributed to inadequate drinking water, sanitation and hygiene [4]. The WHO/UNICEF Joint Monitoring Program for Water Supply, Sanitation and Hygiene (JMP) 2017 report revealed that 844 million people worldwide lack access to basic drinking-water service and 2.3 billion lack basic sanitation services, while 892 million still practiced open defecation [5]. The National Family Health Survey (NFHS-4, 2015–16) reported that in India only 52% of urban households and 18% of rural households have piped water supply, and the main source of water supply among rural households is bore-wells or tube-wells (51%). It has been estimated that 39% of households in India (54% among rural households) have no toilet facility and practicing open defecation [6]. The “indigenous†populations are socially, culturally and economically isolated and usually lack access to basic drinking-water and sanitation services. Therefore, they are vulnerable to ADD outbreaks and other emerging and re-emerging diseases [7]. The United Nations estimates that there are 370 million indigenous people existing across 90 countries of the world. They constitute 5% of the world population but 15% of the poorest [8]. India alone houses more than 705 such indigenous groups termed as Scheduled Tribes. As per the Census 2011, the total Scheduled Tribe population of India is 10.43 crore with a significant proportion of them living in rural areas [9]. On June 26, 2017, Kama-reddy district of Telangana state reported 55 ADD cases from the Pedda-GujjulThanda village. We conducted the outbreak investigation to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods Setting Pedda-Gujjul-Thanda village is a small tribal village with a total population of 563. The village is remotely located as an isolated community with a hilly terrain and is resource-limited with poor accessibility to sanitation and hygiene facilities. The nearest health care facility available for the residents is located at a distance of 10 km from the village. Case definition We defined a case as three or more loose stools within 24 h in a resident of the Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Case finding To find cases, we reviewed medical records of local health care facilities accessed by village residents in the nearby town. We conducted a medical camp in the village during the outbreak period for five days. We conducted a house-to-house survey in the village to find more cases, which are niether reported to health facility nor medical camp. Retrospective cohort study We conducted a retrospective cohort study to identify risk factors associated with illness. We defined the cohort as residents of Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Village resident was the unit of analysis. For data collection, we trained five teams of local paramedical staff. Using a pre-structured questionnaire, we collected data on demographic characteristics and risk factors related to drinking water, sanitation and hygiene. Good hand-washing practice was defined as reported washing of hands with soap and water every time after defecation and before eating. A bore-well less than 30-m-deep, as assessed from the records of village administration, was considered a shallow bore-well. Laboratory and environmental investigations Two stool samples were collected by the treating physician from admitted patients on the first day of hospital admission and transported to the state reference laboratory within two hours in Cary-Blair transport medium. The samples were cultured for Vibrio cholerae, Salmonella and Shigella on nutrient agar, MacConkey agar and deoxycholate citrate agar. Enteric pathogens were identified by biochemical reaction and by agglutination with anti-sera. We collected details of recent rainfall and conducted an environmental survey with household as sampling unit to assess drinking water, sanitation and hygiene practices. We assessed availability of residual chlorine in all village bore-wells and tested four of five bore-wells in the most affected colonies for faecal contamination by H2S method in field. Water was filled up to the ‘fill line’ of the sample bottle and incubated at room temperature (250–370 C) for 36–48 h and observed for colour change in the medium. A water sample was suspected to be contaminated with faecal matter, if it turned black [10, 11]. Because of limited supplies, we were unable to assess the fifth bore-well. Data analysis We analysed the data to describe the occurrence of cases over time, place, and person. We calculated relative risks (RR) with 95% confidence intervals (CI), population attributable risk percentages and conducted multiple logistic regression analysis with the dependent variables including consumption of shallow-downhill bore-well water, report of visible contaminants like mud in drinking water, illiteracy, household water treatment and good hand-washing practice. We used Epi Info version 7.2 for statistical analysis. Results Descriptive epidemiology We identified 191 ADD cases (65% females), with a village attack rate (AR) of 37% (191/512). The attack rate increased with age, with highest among > 60-year age group (55%) and lowest among children under-10 years (11%) (Table 1). No deaths were reported. In addition to diarrhea, cases presented with fever (17%), vomiting (16%) and abdominal pain (13%). 72% (138/191) cases reported to health care facilities and the medical camp conducted in the village. Among the 191 cases, 159 (83%) had mild illness treated with oral rehydration solution; 30 (16%) had moderate dehydration treated with intravenous fluids on out-patient basis, and 2 (1%) with severe dehydration were admitted in the district hospital for treatment with antibiotics (metronidazole and ciprofloxacin) and intravenous fluids. Cases started reported on June 26, 2017, with onset of symptoms from 24 June 2017. Maximum cases were reported on June 27, 2017, and no new cases were reported after June 30, 2017 (Fig. 1). The tribal population in the village had four sub-tribes namely Katroth, Badhawath, Nenawath, Baromath who resided in seven geographically demarcated colonies (labelled as A to G). Katroth sub-tribe resided in colonies A, B and G; Badhawath in colonies C and D; Nenawath in colony E and Baromath in colony F (Table 2). Colonies B and C had higher attack rates (65 and 47% respectively) as compared to other colonies (Fig. 2). Retrospective cohort study Among 563 village residents, 512 (91%) participated in the study. Among the 512 participants, median age was 28 years (range 1–80 years) with 52% females; 50% reported as illiterate with agriculture as the main source of livelihood for 76%. We analysed possible risk factors associated with ADD (Table 3). Drinking water from bore-well groundwater (vs canned water) was found significantly associated with ADD (RR = 12.7; 95% CI = 1.8–87.4). However, only 32 (6%) residents in the village used canned water and bore-well groundwater was the predominant source of water supply. Therefore, we analysed the water sources further, by location and type of bore-wells. Residents who used any of the five shallow bore-wells located downhill were significantly at higher risk (RR = 4.6; 95% CI = 3.4–6.1) and deep bore-wells were protective (RR = 0.4; 95% CI = 0.2–0.9). Report of visible contaminants like mud in drinking water (aOR = 4; 95% CI = 2.1–7.6) and illiteracy (aOR = 3.6; 95% CI = 3.5–10.1) were significantly associated with illness; and household water treatment (done either by boiling or use of candle filters) (aOR = 0.4; 95% CI = 0.2–0.7) and good hand-washing practice (aOR = 0.2; 95% CI = 0.1–0.5) were found protective. Laboratory and environmental results Stool samples collected from two hospitalized cases showed no growth for Vibrio cholerae, Salmonella and Shigella on culture. Among 110 households, 100 (91%) were available for environmental survey. Among the 100 houses surveyed, 79 (79%) were kutcha (low quality) type, made of mud, thatch and other low-quality material. Only 5 (5%) households had a designated toilet at home while the remaining 95 (95%) practiced open defecation at a site located on the slope of the hill behind the downhill colonies B and C (Figs. 2 and 3). Bore-wells were the main source of drinking water supply for 93 (93%) households. There were two deep borewells provided by the village administration and 17 shallow type bore-wells privately constructed by village residents. Five of these 17 (30%) shallow bore-wells were located in colonies B and C, on the downhill slope below the open defecation site. Plastic pipelines from the shallow wells were improperly installed with leakages at multiple points. There was no facility at source, for chlorination or any other mode of purification. Thirty households (30%) treated the water before consumption either by boiling or by use of candle filters. There was no routine drinking water surveillance in place by any authority for assessing the quality and fitness for drinking water. There was no residual chlorine found in any water samples. Three of four drinking water samples from bore-wells of most affected colonies (B and C) indicated faecal contamination by H2S field testing. There was heavy rainfall (average 65 mm in a day) from 22 to 24 June 2017. Prevention and control measures undertaken to contain the outbreak The village residents were discouraged from using shallow bore-well water and were provided with safe canned drinking water until all leakages were secured. Leakages in water supply from the bore-wells were identified and secured. Chlorine tablets were distributed for household level water disinfection. We informed the residents to avoid open defecation near drinking water sources and residential premises. Public health staff conducted health education daily to improve awareness among the villagers about water, sanitation, and hygiene. After active implementation of these control measures, cases declined rapidly in the village (Fig. 1). Discussion A rapid systematic epidemiological investigation of this outbreak identified water contamination points and likely mode of contamination. Based on these findings and our recommendations, the local health department instituted immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families. Effective implementation of public health measures limited the exposure of the community to contaminated water source resulting in rapid containment of the outbreak. Waterborne disease outbreaks tend to have cases spread over a time-period due to ongoing exposure to the contaminated water. In contrast, the pattern of epidemic curve in the present waterborne disease outbreak resembled that of food-borne with a point source exposure. Heavy rains contributed to the run-off of water from the open defecation site into the ground water of shallow wells located on slopes of hilly terrain resulting in heavy contamination and sudden rise of cases. Rapid control measures in the small village, implemented effectively within a short period of time, may have led to rapid decline of cases. The available epidemiological evidence also did not support generation of hypothesis of food-borne origin of the outbreak. In an outbreak reported among school children in Northern Greece in 2012, investigation revealed a waterborne viral gastroenteritis outbreak with a point source pattern, due to consumption of heavily contaminated water from a tap, which was not in use for two weeks during Christmas vacation [12]. Attack rate was high in this outbreak (37%), possibly due to exposure to high pathogen load subsequent to gross faecal contamination of water sources. In the absence of other alternative water sources, this tribal community was exclusively dependent on the contaminated water source for drinking, therefore exposing a large section of the community to risk. Geetha et al. analysed 32 diarrheal outbreaks in south India in non- tribal communities and reported lower attack rates varying from 0.6 to 21.5% [13]. However, tribal populations in India such as in Pedda-Gujjul-Thanda are marginalized with poor availability of WASH facilities [14]. This vulnerable tribal population continues to be at higher risk for ADD outbreaks with 27% not having access to safe drinking water and 75% of households not having toilets [15]. They need special assistance schemes from the government to enable them overcome poor accessibility to WASH facilities and secure healthy living [16]. Due to inadequate availability of communally managed safe public water points by the local authority, this community in Pedda-Gujjul-Thanda village was dependent on privately constructed shallow bore-wells for water supply. These are economical but likely to be unsafe. In this outbreak, open defecation site was present on the downhill slope in proximity to the residential premises and water resources, increasing the risk of drinking water contamination. Among the entire village population, 61% of ADD cases were attributable to drinking water from the ‘shallow downhill bore-wells’ (Population Attributable Fraction 61%), which was also evident from rapid outbreak containment following the elimination of exposure to this single risk factor. Since this exposure factor is amenable to long-term public health intervention, permanent elimination of shallow downhill borewells as water source was recommended, replacing them with properly secured deep bore-wells. Shallow bore-wells are known for their susceptibility to contamination from surface land-use activities [17, 18]. Studies have found levels of E. coli and enteric viruses to be high in shallow sources of ground water especially when they are in close proximity to polluting sources [19–21]. Consumption of ground water from shallow bore-wells with no purification facility increases the risk of diarrhea outbreaks manifold [22, 23]. A metaanalytic study of water-borne diarrheal disease outbreaks in China reported that 78 of 85 (92%) outbreaks (between year 1987 to 2014) were due to poor sanitary conditions of wells with lavatories/septic tanks nearby and lack of purification facilities [24]. In developed countries and urban areas of developing countries, as water supply and sanitation have improved dramatically over a period of time, such outbreaks were rarely reported in the recent past. The largest E. coli O157 outbreak in United States occurred in 1999 at a county fair (781 ill persons and 2 deaths) was due to groundwater source from a temporary unregulated well at the fairground [25]. Our findings have implication for India’s progress towards United Nation’s Sustainable Development Goal (SDG) 6 and India’s nation-wide campaign ‘Swachh Bharat Mission (SBM)’ to ensure availability and management of water and sanitation for all. SDG 6 aims at achieving universal access to basic sanitation service by 2030; and it has been reported that between 2000 and 2017, the proportion lacking even a basic sanitation service decreased from 44 to 27% [26]. SBM aims to achieve an open-defecation free status in rural areas through the construction of household-owned and community-owned toilets and establishing an accountable mechanism of monitoring toilet use. In 2015 in India, around 524 million (39%) practiced open defecation. However, under the SBM mission, due to increase in ‘households with toilets’ only 19 million (1.4%) practiced open defecation in January 2019 [5, 27]. There has also been a 71.58% increase in ‘households with toilets’ from October 2014 to October 2019 in rural areas of the Telangana state in India [27]. The tribal community initially obstructed the effective delivery of health care services; however, after involvement of the local stakeholders and tribal leaders, the acceptance towards medical treatment and community health services improved. Notwithstanding, most of the patients were still reluctant and did not consent for giving stool specimens for laboratory diagnosis. Establishing a rapport with the reticent tribal community was a major challenge faced by the outbreak investigation team. Lack of microbiological aetiology confirmation of the outbreak remained a limitation of the investigation due to limited stool samples and laboratory-capacity constraints of the remote area. Recognizing the pivotal importance of SDGs, national health policy of India (2017) has set the health-related cross-sectorial goal “access to safe water and sanitation to all by 2020†[28]. Greater political and financial commitment towards resource-limited remote tribal areas with effective community mobilization is required to accelerate the public health interventions to improve WASH and to prevent ADD outbreaks in the future. Conclusion This was a community-wide acute diarrheal disease outbreak with high village attack rate in a remote tribal village of Telangana with poor availability of safe water, sanitation and hygiene (WASH) facilities. A rapid and systematic epidemiological investigation identified drinking of faecal-contaminated water from the shallow borewells as the leading cause for this outbreak. These borewells were likely contaminated from runoff after rain from open defecation areas located on a downhill slope. Prompt and targeted public health action contained the number of cases.

In which city did this happen?

89

A community-wide acute diarrheal disease outbreak associated with drinking contaminated water from shallow bore-wells in a tribal village

Abstract Background: In 2016, India reported 709 acute diarrheal disease (ADD) outbreaks (> 25% of all outbreaks). Tribal populations are at higher risk with 27% not having accessibility to safe drinking water and 75% households not having toilets. On June 26, 2017 Pedda-Gujjul-Thanda, a tribal village reported an acute diarrheal disease (ADD) outbreak. We investigated to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods: We defined a case as ≥3 loose stools within 24 h in Pedda-Gujjul-Thanda residents from June 24–30, 2017. We identified cases by reviewing hospital records and house-to-house survey. We conducted a retrospective cohort study and collected stool samples for culture. We assessed drinking water supply and sanitation practices and tested water samples for faecal-contamination. Results: We identified 191 cases (65% females) with median age 36 years (range 4–80 years) and no deaths. The attack-rate (AR) was 37% (191/512). Downhill colonies (located on slope of hilly terrains of the village) reported higher ARs (56%[136/243], p < 0.001) than others (20%[55/269]). Symptoms included diarrhea (100%), fever (17%), vomiting (16%) and abdominal pain (13%). Drinking water from five shallow bore-wells located in downhill colonies was significantly associated with illness (RR = 4.6, 95%CI = 3.4–6.1 and population attributable fraction 61%). In multivariate analysis, drinking water from the shallow bore-wells located in downhill colonies (aOR = 7.9, [95% CI =4.7– 13.2]), illiteracy (aOR =6, [95% CI = 3.6–10.1]), good hand-washing practice (aOR = 0.4, [95%CI = 0.2–0.7]) and household water treatment (aOR = 0.3, [95%CI = 0.2–0.5]) were significantly associated with illness. Two stool cultures were negative for Vibrio cholerae. Heavy rainfall was reported from June 22–24. Five of six water samples collected from shallow bore-wells located in downhill colonies were positive for faecal contamination. Conclusion: An ADD outbreak with high attack rate in a remote tribal village was associated with drinking water from shallow downhill bore-wells, likely contaminated via runoff from open defecation areas after heavy rains. Based on our recommendations, immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families, and as long-term public health measures construction of house-hold latrines and piped-water supply initiated. Keywords: Acute diarrheal disease, Outbreak, Bore-well, Tribal Introduction Globally there are an estimated 1.7 billion cases and 2.2 million deaths from acute diarrheal disease (ADD) every year [1]. In India, the burden is particularly high with more than 13.9 million cases reported in 2016 and 709 ADD outbreaks reported accounting to more than 25% of all outbreaks [2, 3]. Lack of access to safe drinking water and basic sanitation are the leading causes of ADD burden globally and in India. It is estimated that globally 58% of ADD deaths are attributed to inadequate drinking water, sanitation and hygiene [4]. The WHO/UNICEF Joint Monitoring Program for Water Supply, Sanitation and Hygiene (JMP) 2017 report revealed that 844 million people worldwide lack access to basic drinking-water service and 2.3 billion lack basic sanitation services, while 892 million still practiced open defecation [5]. The National Family Health Survey (NFHS-4, 2015–16) reported that in India only 52% of urban households and 18% of rural households have piped water supply, and the main source of water supply among rural households is bore-wells or tube-wells (51%). It has been estimated that 39% of households in India (54% among rural households) have no toilet facility and practicing open defecation [6]. The “indigenous†populations are socially, culturally and economically isolated and usually lack access to basic drinking-water and sanitation services. Therefore, they are vulnerable to ADD outbreaks and other emerging and re-emerging diseases [7]. The United Nations estimates that there are 370 million indigenous people existing across 90 countries of the world. They constitute 5% of the world population but 15% of the poorest [8]. India alone houses more than 705 such indigenous groups termed as Scheduled Tribes. As per the Census 2011, the total Scheduled Tribe population of India is 10.43 crore with a significant proportion of them living in rural areas [9]. On June 26, 2017, Kama-reddy district of Telangana state reported 55 ADD cases from the Pedda-GujjulThanda village. We conducted the outbreak investigation to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods Setting Pedda-Gujjul-Thanda village is a small tribal village with a total population of 563. The village is remotely located as an isolated community with a hilly terrain and is resource-limited with poor accessibility to sanitation and hygiene facilities. The nearest health care facility available for the residents is located at a distance of 10 km from the village. Case definition We defined a case as three or more loose stools within 24 h in a resident of the Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Case finding To find cases, we reviewed medical records of local health care facilities accessed by village residents in the nearby town. We conducted a medical camp in the village during the outbreak period for five days. We conducted a house-to-house survey in the village to find more cases, which are niether reported to health facility nor medical camp. Retrospective cohort study We conducted a retrospective cohort study to identify risk factors associated with illness. We defined the cohort as residents of Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Village resident was the unit of analysis. For data collection, we trained five teams of local paramedical staff. Using a pre-structured questionnaire, we collected data on demographic characteristics and risk factors related to drinking water, sanitation and hygiene. Good hand-washing practice was defined as reported washing of hands with soap and water every time after defecation and before eating. A bore-well less than 30-m-deep, as assessed from the records of village administration, was considered a shallow bore-well. Laboratory and environmental investigations Two stool samples were collected by the treating physician from admitted patients on the first day of hospital admission and transported to the state reference laboratory within two hours in Cary-Blair transport medium. The samples were cultured for Vibrio cholerae, Salmonella and Shigella on nutrient agar, MacConkey agar and deoxycholate citrate agar. Enteric pathogens were identified by biochemical reaction and by agglutination with anti-sera. We collected details of recent rainfall and conducted an environmental survey with household as sampling unit to assess drinking water, sanitation and hygiene practices. We assessed availability of residual chlorine in all village bore-wells and tested four of five bore-wells in the most affected colonies for faecal contamination by H2S method in field. Water was filled up to the ‘fill line’ of the sample bottle and incubated at room temperature (250–370 C) for 36–48 h and observed for colour change in the medium. A water sample was suspected to be contaminated with faecal matter, if it turned black [10, 11]. Because of limited supplies, we were unable to assess the fifth bore-well. Data analysis We analysed the data to describe the occurrence of cases over time, place, and person. We calculated relative risks (RR) with 95% confidence intervals (CI), population attributable risk percentages and conducted multiple logistic regression analysis with the dependent variables including consumption of shallow-downhill bore-well water, report of visible contaminants like mud in drinking water, illiteracy, household water treatment and good hand-washing practice. We used Epi Info version 7.2 for statistical analysis. Results Descriptive epidemiology We identified 191 ADD cases (65% females), with a village attack rate (AR) of 37% (191/512). The attack rate increased with age, with highest among > 60-year age group (55%) and lowest among children under-10 years (11%) (Table 1). No deaths were reported. In addition to diarrhea, cases presented with fever (17%), vomiting (16%) and abdominal pain (13%). 72% (138/191) cases reported to health care facilities and the medical camp conducted in the village. Among the 191 cases, 159 (83%) had mild illness treated with oral rehydration solution; 30 (16%) had moderate dehydration treated with intravenous fluids on out-patient basis, and 2 (1%) with severe dehydration were admitted in the district hospital for treatment with antibiotics (metronidazole and ciprofloxacin) and intravenous fluids. Cases started reported on June 26, 2017, with onset of symptoms from 24 June 2017. Maximum cases were reported on June 27, 2017, and no new cases were reported after June 30, 2017 (Fig. 1). The tribal population in the village had four sub-tribes namely Katroth, Badhawath, Nenawath, Baromath who resided in seven geographically demarcated colonies (labelled as A to G). Katroth sub-tribe resided in colonies A, B and G; Badhawath in colonies C and D; Nenawath in colony E and Baromath in colony F (Table 2). Colonies B and C had higher attack rates (65 and 47% respectively) as compared to other colonies (Fig. 2). Retrospective cohort study Among 563 village residents, 512 (91%) participated in the study. Among the 512 participants, median age was 28 years (range 1–80 years) with 52% females; 50% reported as illiterate with agriculture as the main source of livelihood for 76%. We analysed possible risk factors associated with ADD (Table 3). Drinking water from bore-well groundwater (vs canned water) was found significantly associated with ADD (RR = 12.7; 95% CI = 1.8–87.4). However, only 32 (6%) residents in the village used canned water and bore-well groundwater was the predominant source of water supply. Therefore, we analysed the water sources further, by location and type of bore-wells. Residents who used any of the five shallow bore-wells located downhill were significantly at higher risk (RR = 4.6; 95% CI = 3.4–6.1) and deep bore-wells were protective (RR = 0.4; 95% CI = 0.2–0.9). Report of visible contaminants like mud in drinking water (aOR = 4; 95% CI = 2.1–7.6) and illiteracy (aOR = 3.6; 95% CI = 3.5–10.1) were significantly associated with illness; and household water treatment (done either by boiling or use of candle filters) (aOR = 0.4; 95% CI = 0.2–0.7) and good hand-washing practice (aOR = 0.2; 95% CI = 0.1–0.5) were found protective. Laboratory and environmental results Stool samples collected from two hospitalized cases showed no growth for Vibrio cholerae, Salmonella and Shigella on culture. Among 110 households, 100 (91%) were available for environmental survey. Among the 100 houses surveyed, 79 (79%) were kutcha (low quality) type, made of mud, thatch and other low-quality material. Only 5 (5%) households had a designated toilet at home while the remaining 95 (95%) practiced open defecation at a site located on the slope of the hill behind the downhill colonies B and C (Figs. 2 and 3). Bore-wells were the main source of drinking water supply for 93 (93%) households. There were two deep borewells provided by the village administration and 17 shallow type bore-wells privately constructed by village residents. Five of these 17 (30%) shallow bore-wells were located in colonies B and C, on the downhill slope below the open defecation site. Plastic pipelines from the shallow wells were improperly installed with leakages at multiple points. There was no facility at source, for chlorination or any other mode of purification. Thirty households (30%) treated the water before consumption either by boiling or by use of candle filters. There was no routine drinking water surveillance in place by any authority for assessing the quality and fitness for drinking water. There was no residual chlorine found in any water samples. Three of four drinking water samples from bore-wells of most affected colonies (B and C) indicated faecal contamination by H2S field testing. There was heavy rainfall (average 65 mm in a day) from 22 to 24 June 2017. Prevention and control measures undertaken to contain the outbreak The village residents were discouraged from using shallow bore-well water and were provided with safe canned drinking water until all leakages were secured. Leakages in water supply from the bore-wells were identified and secured. Chlorine tablets were distributed for household level water disinfection. We informed the residents to avoid open defecation near drinking water sources and residential premises. Public health staff conducted health education daily to improve awareness among the villagers about water, sanitation, and hygiene. After active implementation of these control measures, cases declined rapidly in the village (Fig. 1). Discussion A rapid systematic epidemiological investigation of this outbreak identified water contamination points and likely mode of contamination. Based on these findings and our recommendations, the local health department instituted immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families. Effective implementation of public health measures limited the exposure of the community to contaminated water source resulting in rapid containment of the outbreak. Waterborne disease outbreaks tend to have cases spread over a time-period due to ongoing exposure to the contaminated water. In contrast, the pattern of epidemic curve in the present waterborne disease outbreak resembled that of food-borne with a point source exposure. Heavy rains contributed to the run-off of water from the open defecation site into the ground water of shallow wells located on slopes of hilly terrain resulting in heavy contamination and sudden rise of cases. Rapid control measures in the small village, implemented effectively within a short period of time, may have led to rapid decline of cases. The available epidemiological evidence also did not support generation of hypothesis of food-borne origin of the outbreak. In an outbreak reported among school children in Northern Greece in 2012, investigation revealed a waterborne viral gastroenteritis outbreak with a point source pattern, due to consumption of heavily contaminated water from a tap, which was not in use for two weeks during Christmas vacation [12]. Attack rate was high in this outbreak (37%), possibly due to exposure to high pathogen load subsequent to gross faecal contamination of water sources. In the absence of other alternative water sources, this tribal community was exclusively dependent on the contaminated water source for drinking, therefore exposing a large section of the community to risk. Geetha et al. analysed 32 diarrheal outbreaks in south India in non- tribal communities and reported lower attack rates varying from 0.6 to 21.5% [13]. However, tribal populations in India such as in Pedda-Gujjul-Thanda are marginalized with poor availability of WASH facilities [14]. This vulnerable tribal population continues to be at higher risk for ADD outbreaks with 27% not having access to safe drinking water and 75% of households not having toilets [15]. They need special assistance schemes from the government to enable them overcome poor accessibility to WASH facilities and secure healthy living [16]. Due to inadequate availability of communally managed safe public water points by the local authority, this community in Pedda-Gujjul-Thanda village was dependent on privately constructed shallow bore-wells for water supply. These are economical but likely to be unsafe. In this outbreak, open defecation site was present on the downhill slope in proximity to the residential premises and water resources, increasing the risk of drinking water contamination. Among the entire village population, 61% of ADD cases were attributable to drinking water from the ‘shallow downhill bore-wells’ (Population Attributable Fraction 61%), which was also evident from rapid outbreak containment following the elimination of exposure to this single risk factor. Since this exposure factor is amenable to long-term public health intervention, permanent elimination of shallow downhill borewells as water source was recommended, replacing them with properly secured deep bore-wells. Shallow bore-wells are known for their susceptibility to contamination from surface land-use activities [17, 18]. Studies have found levels of E. coli and enteric viruses to be high in shallow sources of ground water especially when they are in close proximity to polluting sources [19–21]. Consumption of ground water from shallow bore-wells with no purification facility increases the risk of diarrhea outbreaks manifold [22, 23]. A metaanalytic study of water-borne diarrheal disease outbreaks in China reported that 78 of 85 (92%) outbreaks (between year 1987 to 2014) were due to poor sanitary conditions of wells with lavatories/septic tanks nearby and lack of purification facilities [24]. In developed countries and urban areas of developing countries, as water supply and sanitation have improved dramatically over a period of time, such outbreaks were rarely reported in the recent past. The largest E. coli O157 outbreak in United States occurred in 1999 at a county fair (781 ill persons and 2 deaths) was due to groundwater source from a temporary unregulated well at the fairground [25]. Our findings have implication for India’s progress towards United Nation’s Sustainable Development Goal (SDG) 6 and India’s nation-wide campaign ‘Swachh Bharat Mission (SBM)’ to ensure availability and management of water and sanitation for all. SDG 6 aims at achieving universal access to basic sanitation service by 2030; and it has been reported that between 2000 and 2017, the proportion lacking even a basic sanitation service decreased from 44 to 27% [26]. SBM aims to achieve an open-defecation free status in rural areas through the construction of household-owned and community-owned toilets and establishing an accountable mechanism of monitoring toilet use. In 2015 in India, around 524 million (39%) practiced open defecation. However, under the SBM mission, due to increase in ‘households with toilets’ only 19 million (1.4%) practiced open defecation in January 2019 [5, 27]. There has also been a 71.58% increase in ‘households with toilets’ from October 2014 to October 2019 in rural areas of the Telangana state in India [27]. The tribal community initially obstructed the effective delivery of health care services; however, after involvement of the local stakeholders and tribal leaders, the acceptance towards medical treatment and community health services improved. Notwithstanding, most of the patients were still reluctant and did not consent for giving stool specimens for laboratory diagnosis. Establishing a rapport with the reticent tribal community was a major challenge faced by the outbreak investigation team. Lack of microbiological aetiology confirmation of the outbreak remained a limitation of the investigation due to limited stool samples and laboratory-capacity constraints of the remote area. Recognizing the pivotal importance of SDGs, national health policy of India (2017) has set the health-related cross-sectorial goal “access to safe water and sanitation to all by 2020†[28]. Greater political and financial commitment towards resource-limited remote tribal areas with effective community mobilization is required to accelerate the public health interventions to improve WASH and to prevent ADD outbreaks in the future. Conclusion This was a community-wide acute diarrheal disease outbreak with high village attack rate in a remote tribal village of Telangana with poor availability of safe water, sanitation and hygiene (WASH) facilities. A rapid and systematic epidemiological investigation identified drinking of faecal-contaminated water from the shallow borewells as the leading cause for this outbreak. These borewells were likely contaminated from runoff after rain from open defecation areas located on a downhill slope. Prompt and targeted public health action contained the number of cases.

In which region did this happen?

90

A community-wide acute diarrheal disease outbreak associated with drinking contaminated water from shallow bore-wells in a tribal village

Abstract Background: In 2016, India reported 709 acute diarrheal disease (ADD) outbreaks (> 25% of all outbreaks). Tribal populations are at higher risk with 27% not having accessibility to safe drinking water and 75% households not having toilets. On June 26, 2017 Pedda-Gujjul-Thanda, a tribal village reported an acute diarrheal disease (ADD) outbreak. We investigated to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods: We defined a case as ≥3 loose stools within 24 h in Pedda-Gujjul-Thanda residents from June 24–30, 2017. We identified cases by reviewing hospital records and house-to-house survey. We conducted a retrospective cohort study and collected stool samples for culture. We assessed drinking water supply and sanitation practices and tested water samples for faecal-contamination. Results: We identified 191 cases (65% females) with median age 36 years (range 4–80 years) and no deaths. The attack-rate (AR) was 37% (191/512). Downhill colonies (located on slope of hilly terrains of the village) reported higher ARs (56%[136/243], p < 0.001) than others (20%[55/269]). Symptoms included diarrhea (100%), fever (17%), vomiting (16%) and abdominal pain (13%). Drinking water from five shallow bore-wells located in downhill colonies was significantly associated with illness (RR = 4.6, 95%CI = 3.4–6.1 and population attributable fraction 61%). In multivariate analysis, drinking water from the shallow bore-wells located in downhill colonies (aOR = 7.9, [95% CI =4.7– 13.2]), illiteracy (aOR =6, [95% CI = 3.6–10.1]), good hand-washing practice (aOR = 0.4, [95%CI = 0.2–0.7]) and household water treatment (aOR = 0.3, [95%CI = 0.2–0.5]) were significantly associated with illness. Two stool cultures were negative for Vibrio cholerae. Heavy rainfall was reported from June 22–24. Five of six water samples collected from shallow bore-wells located in downhill colonies were positive for faecal contamination. Conclusion: An ADD outbreak with high attack rate in a remote tribal village was associated with drinking water from shallow downhill bore-wells, likely contaminated via runoff from open defecation areas after heavy rains. Based on our recommendations, immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families, and as long-term public health measures construction of house-hold latrines and piped-water supply initiated. Keywords: Acute diarrheal disease, Outbreak, Bore-well, Tribal Introduction Globally there are an estimated 1.7 billion cases and 2.2 million deaths from acute diarrheal disease (ADD) every year [1]. In India, the burden is particularly high with more than 13.9 million cases reported in 2016 and 709 ADD outbreaks reported accounting to more than 25% of all outbreaks [2, 3]. Lack of access to safe drinking water and basic sanitation are the leading causes of ADD burden globally and in India. It is estimated that globally 58% of ADD deaths are attributed to inadequate drinking water, sanitation and hygiene [4]. The WHO/UNICEF Joint Monitoring Program for Water Supply, Sanitation and Hygiene (JMP) 2017 report revealed that 844 million people worldwide lack access to basic drinking-water service and 2.3 billion lack basic sanitation services, while 892 million still practiced open defecation [5]. The National Family Health Survey (NFHS-4, 2015–16) reported that in India only 52% of urban households and 18% of rural households have piped water supply, and the main source of water supply among rural households is bore-wells or tube-wells (51%). It has been estimated that 39% of households in India (54% among rural households) have no toilet facility and practicing open defecation [6]. The “indigenous†populations are socially, culturally and economically isolated and usually lack access to basic drinking-water and sanitation services. Therefore, they are vulnerable to ADD outbreaks and other emerging and re-emerging diseases [7]. The United Nations estimates that there are 370 million indigenous people existing across 90 countries of the world. They constitute 5% of the world population but 15% of the poorest [8]. India alone houses more than 705 such indigenous groups termed as Scheduled Tribes. As per the Census 2011, the total Scheduled Tribe population of India is 10.43 crore with a significant proportion of them living in rural areas [9]. On June 26, 2017, Kama-reddy district of Telangana state reported 55 ADD cases from the Pedda-GujjulThanda village. We conducted the outbreak investigation to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods Setting Pedda-Gujjul-Thanda village is a small tribal village with a total population of 563. The village is remotely located as an isolated community with a hilly terrain and is resource-limited with poor accessibility to sanitation and hygiene facilities. The nearest health care facility available for the residents is located at a distance of 10 km from the village. Case definition We defined a case as three or more loose stools within 24 h in a resident of the Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Case finding To find cases, we reviewed medical records of local health care facilities accessed by village residents in the nearby town. We conducted a medical camp in the village during the outbreak period for five days. We conducted a house-to-house survey in the village to find more cases, which are niether reported to health facility nor medical camp. Retrospective cohort study We conducted a retrospective cohort study to identify risk factors associated with illness. We defined the cohort as residents of Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Village resident was the unit of analysis. For data collection, we trained five teams of local paramedical staff. Using a pre-structured questionnaire, we collected data on demographic characteristics and risk factors related to drinking water, sanitation and hygiene. Good hand-washing practice was defined as reported washing of hands with soap and water every time after defecation and before eating. A bore-well less than 30-m-deep, as assessed from the records of village administration, was considered a shallow bore-well. Laboratory and environmental investigations Two stool samples were collected by the treating physician from admitted patients on the first day of hospital admission and transported to the state reference laboratory within two hours in Cary-Blair transport medium. The samples were cultured for Vibrio cholerae, Salmonella and Shigella on nutrient agar, MacConkey agar and deoxycholate citrate agar. Enteric pathogens were identified by biochemical reaction and by agglutination with anti-sera. We collected details of recent rainfall and conducted an environmental survey with household as sampling unit to assess drinking water, sanitation and hygiene practices. We assessed availability of residual chlorine in all village bore-wells and tested four of five bore-wells in the most affected colonies for faecal contamination by H2S method in field. Water was filled up to the ‘fill line’ of the sample bottle and incubated at room temperature (250–370 C) for 36–48 h and observed for colour change in the medium. A water sample was suspected to be contaminated with faecal matter, if it turned black [10, 11]. Because of limited supplies, we were unable to assess the fifth bore-well. Data analysis We analysed the data to describe the occurrence of cases over time, place, and person. We calculated relative risks (RR) with 95% confidence intervals (CI), population attributable risk percentages and conducted multiple logistic regression analysis with the dependent variables including consumption of shallow-downhill bore-well water, report of visible contaminants like mud in drinking water, illiteracy, household water treatment and good hand-washing practice. We used Epi Info version 7.2 for statistical analysis. Results Descriptive epidemiology We identified 191 ADD cases (65% females), with a village attack rate (AR) of 37% (191/512). The attack rate increased with age, with highest among > 60-year age group (55%) and lowest among children under-10 years (11%) (Table 1). No deaths were reported. In addition to diarrhea, cases presented with fever (17%), vomiting (16%) and abdominal pain (13%). 72% (138/191) cases reported to health care facilities and the medical camp conducted in the village. Among the 191 cases, 159 (83%) had mild illness treated with oral rehydration solution; 30 (16%) had moderate dehydration treated with intravenous fluids on out-patient basis, and 2 (1%) with severe dehydration were admitted in the district hospital for treatment with antibiotics (metronidazole and ciprofloxacin) and intravenous fluids. Cases started reported on June 26, 2017, with onset of symptoms from 24 June 2017. Maximum cases were reported on June 27, 2017, and no new cases were reported after June 30, 2017 (Fig. 1). The tribal population in the village had four sub-tribes namely Katroth, Badhawath, Nenawath, Baromath who resided in seven geographically demarcated colonies (labelled as A to G). Katroth sub-tribe resided in colonies A, B and G; Badhawath in colonies C and D; Nenawath in colony E and Baromath in colony F (Table 2). Colonies B and C had higher attack rates (65 and 47% respectively) as compared to other colonies (Fig. 2). Retrospective cohort study Among 563 village residents, 512 (91%) participated in the study. Among the 512 participants, median age was 28 years (range 1–80 years) with 52% females; 50% reported as illiterate with agriculture as the main source of livelihood for 76%. We analysed possible risk factors associated with ADD (Table 3). Drinking water from bore-well groundwater (vs canned water) was found significantly associated with ADD (RR = 12.7; 95% CI = 1.8–87.4). However, only 32 (6%) residents in the village used canned water and bore-well groundwater was the predominant source of water supply. Therefore, we analysed the water sources further, by location and type of bore-wells. Residents who used any of the five shallow bore-wells located downhill were significantly at higher risk (RR = 4.6; 95% CI = 3.4–6.1) and deep bore-wells were protective (RR = 0.4; 95% CI = 0.2–0.9). Report of visible contaminants like mud in drinking water (aOR = 4; 95% CI = 2.1–7.6) and illiteracy (aOR = 3.6; 95% CI = 3.5–10.1) were significantly associated with illness; and household water treatment (done either by boiling or use of candle filters) (aOR = 0.4; 95% CI = 0.2–0.7) and good hand-washing practice (aOR = 0.2; 95% CI = 0.1–0.5) were found protective. Laboratory and environmental results Stool samples collected from two hospitalized cases showed no growth for Vibrio cholerae, Salmonella and Shigella on culture. Among 110 households, 100 (91%) were available for environmental survey. Among the 100 houses surveyed, 79 (79%) were kutcha (low quality) type, made of mud, thatch and other low-quality material. Only 5 (5%) households had a designated toilet at home while the remaining 95 (95%) practiced open defecation at a site located on the slope of the hill behind the downhill colonies B and C (Figs. 2 and 3). Bore-wells were the main source of drinking water supply for 93 (93%) households. There were two deep borewells provided by the village administration and 17 shallow type bore-wells privately constructed by village residents. Five of these 17 (30%) shallow bore-wells were located in colonies B and C, on the downhill slope below the open defecation site. Plastic pipelines from the shallow wells were improperly installed with leakages at multiple points. There was no facility at source, for chlorination or any other mode of purification. Thirty households (30%) treated the water before consumption either by boiling or by use of candle filters. There was no routine drinking water surveillance in place by any authority for assessing the quality and fitness for drinking water. There was no residual chlorine found in any water samples. Three of four drinking water samples from bore-wells of most affected colonies (B and C) indicated faecal contamination by H2S field testing. There was heavy rainfall (average 65 mm in a day) from 22 to 24 June 2017. Prevention and control measures undertaken to contain the outbreak The village residents were discouraged from using shallow bore-well water and were provided with safe canned drinking water until all leakages were secured. Leakages in water supply from the bore-wells were identified and secured. Chlorine tablets were distributed for household level water disinfection. We informed the residents to avoid open defecation near drinking water sources and residential premises. Public health staff conducted health education daily to improve awareness among the villagers about water, sanitation, and hygiene. After active implementation of these control measures, cases declined rapidly in the village (Fig. 1). Discussion A rapid systematic epidemiological investigation of this outbreak identified water contamination points and likely mode of contamination. Based on these findings and our recommendations, the local health department instituted immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families. Effective implementation of public health measures limited the exposure of the community to contaminated water source resulting in rapid containment of the outbreak. Waterborne disease outbreaks tend to have cases spread over a time-period due to ongoing exposure to the contaminated water. In contrast, the pattern of epidemic curve in the present waterborne disease outbreak resembled that of food-borne with a point source exposure. Heavy rains contributed to the run-off of water from the open defecation site into the ground water of shallow wells located on slopes of hilly terrain resulting in heavy contamination and sudden rise of cases. Rapid control measures in the small village, implemented effectively within a short period of time, may have led to rapid decline of cases. The available epidemiological evidence also did not support generation of hypothesis of food-borne origin of the outbreak. In an outbreak reported among school children in Northern Greece in 2012, investigation revealed a waterborne viral gastroenteritis outbreak with a point source pattern, due to consumption of heavily contaminated water from a tap, which was not in use for two weeks during Christmas vacation [12]. Attack rate was high in this outbreak (37%), possibly due to exposure to high pathogen load subsequent to gross faecal contamination of water sources. In the absence of other alternative water sources, this tribal community was exclusively dependent on the contaminated water source for drinking, therefore exposing a large section of the community to risk. Geetha et al. analysed 32 diarrheal outbreaks in south India in non- tribal communities and reported lower attack rates varying from 0.6 to 21.5% [13]. However, tribal populations in India such as in Pedda-Gujjul-Thanda are marginalized with poor availability of WASH facilities [14]. This vulnerable tribal population continues to be at higher risk for ADD outbreaks with 27% not having access to safe drinking water and 75% of households not having toilets [15]. They need special assistance schemes from the government to enable them overcome poor accessibility to WASH facilities and secure healthy living [16]. Due to inadequate availability of communally managed safe public water points by the local authority, this community in Pedda-Gujjul-Thanda village was dependent on privately constructed shallow bore-wells for water supply. These are economical but likely to be unsafe. In this outbreak, open defecation site was present on the downhill slope in proximity to the residential premises and water resources, increasing the risk of drinking water contamination. Among the entire village population, 61% of ADD cases were attributable to drinking water from the ‘shallow downhill bore-wells’ (Population Attributable Fraction 61%), which was also evident from rapid outbreak containment following the elimination of exposure to this single risk factor. Since this exposure factor is amenable to long-term public health intervention, permanent elimination of shallow downhill borewells as water source was recommended, replacing them with properly secured deep bore-wells. Shallow bore-wells are known for their susceptibility to contamination from surface land-use activities [17, 18]. Studies have found levels of E. coli and enteric viruses to be high in shallow sources of ground water especially when they are in close proximity to polluting sources [19–21]. Consumption of ground water from shallow bore-wells with no purification facility increases the risk of diarrhea outbreaks manifold [22, 23]. A metaanalytic study of water-borne diarrheal disease outbreaks in China reported that 78 of 85 (92%) outbreaks (between year 1987 to 2014) were due to poor sanitary conditions of wells with lavatories/septic tanks nearby and lack of purification facilities [24]. In developed countries and urban areas of developing countries, as water supply and sanitation have improved dramatically over a period of time, such outbreaks were rarely reported in the recent past. The largest E. coli O157 outbreak in United States occurred in 1999 at a county fair (781 ill persons and 2 deaths) was due to groundwater source from a temporary unregulated well at the fairground [25]. Our findings have implication for India’s progress towards United Nation’s Sustainable Development Goal (SDG) 6 and India’s nation-wide campaign ‘Swachh Bharat Mission (SBM)’ to ensure availability and management of water and sanitation for all. SDG 6 aims at achieving universal access to basic sanitation service by 2030; and it has been reported that between 2000 and 2017, the proportion lacking even a basic sanitation service decreased from 44 to 27% [26]. SBM aims to achieve an open-defecation free status in rural areas through the construction of household-owned and community-owned toilets and establishing an accountable mechanism of monitoring toilet use. In 2015 in India, around 524 million (39%) practiced open defecation. However, under the SBM mission, due to increase in ‘households with toilets’ only 19 million (1.4%) practiced open defecation in January 2019 [5, 27]. There has also been a 71.58% increase in ‘households with toilets’ from October 2014 to October 2019 in rural areas of the Telangana state in India [27]. The tribal community initially obstructed the effective delivery of health care services; however, after involvement of the local stakeholders and tribal leaders, the acceptance towards medical treatment and community health services improved. Notwithstanding, most of the patients were still reluctant and did not consent for giving stool specimens for laboratory diagnosis. Establishing a rapport with the reticent tribal community was a major challenge faced by the outbreak investigation team. Lack of microbiological aetiology confirmation of the outbreak remained a limitation of the investigation due to limited stool samples and laboratory-capacity constraints of the remote area. Recognizing the pivotal importance of SDGs, national health policy of India (2017) has set the health-related cross-sectorial goal “access to safe water and sanitation to all by 2020†[28]. Greater political and financial commitment towards resource-limited remote tribal areas with effective community mobilization is required to accelerate the public health interventions to improve WASH and to prevent ADD outbreaks in the future. Conclusion This was a community-wide acute diarrheal disease outbreak with high village attack rate in a remote tribal village of Telangana with poor availability of safe water, sanitation and hygiene (WASH) facilities. A rapid and systematic epidemiological investigation identified drinking of faecal-contaminated water from the shallow borewells as the leading cause for this outbreak. These borewells were likely contaminated from runoff after rain from open defecation areas located on a downhill slope. Prompt and targeted public health action contained the number of cases.

In which country did this happen?

91

A community-wide acute diarrheal disease outbreak associated with drinking contaminated water from shallow bore-wells in a tribal village

Abstract Background: In 2016, India reported 709 acute diarrheal disease (ADD) outbreaks (> 25% of all outbreaks). Tribal populations are at higher risk with 27% not having accessibility to safe drinking water and 75% households not having toilets. On June 26, 2017 Pedda-Gujjul-Thanda, a tribal village reported an acute diarrheal disease (ADD) outbreak. We investigated to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods: We defined a case as ≥3 loose stools within 24 h in Pedda-Gujjul-Thanda residents from June 24–30, 2017. We identified cases by reviewing hospital records and house-to-house survey. We conducted a retrospective cohort study and collected stool samples for culture. We assessed drinking water supply and sanitation practices and tested water samples for faecal-contamination. Results: We identified 191 cases (65% females) with median age 36 years (range 4–80 years) and no deaths. The attack-rate (AR) was 37% (191/512). Downhill colonies (located on slope of hilly terrains of the village) reported higher ARs (56%[136/243], p < 0.001) than others (20%[55/269]). Symptoms included diarrhea (100%), fever (17%), vomiting (16%) and abdominal pain (13%). Drinking water from five shallow bore-wells located in downhill colonies was significantly associated with illness (RR = 4.6, 95%CI = 3.4–6.1 and population attributable fraction 61%). In multivariate analysis, drinking water from the shallow bore-wells located in downhill colonies (aOR = 7.9, [95% CI =4.7– 13.2]), illiteracy (aOR =6, [95% CI = 3.6–10.1]), good hand-washing practice (aOR = 0.4, [95%CI = 0.2–0.7]) and household water treatment (aOR = 0.3, [95%CI = 0.2–0.5]) were significantly associated with illness. Two stool cultures were negative for Vibrio cholerae. Heavy rainfall was reported from June 22–24. Five of six water samples collected from shallow bore-wells located in downhill colonies were positive for faecal contamination. Conclusion: An ADD outbreak with high attack rate in a remote tribal village was associated with drinking water from shallow downhill bore-wells, likely contaminated via runoff from open defecation areas after heavy rains. Based on our recommendations, immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families, and as long-term public health measures construction of house-hold latrines and piped-water supply initiated. Keywords: Acute diarrheal disease, Outbreak, Bore-well, Tribal Introduction Globally there are an estimated 1.7 billion cases and 2.2 million deaths from acute diarrheal disease (ADD) every year [1]. In India, the burden is particularly high with more than 13.9 million cases reported in 2016 and 709 ADD outbreaks reported accounting to more than 25% of all outbreaks [2, 3]. Lack of access to safe drinking water and basic sanitation are the leading causes of ADD burden globally and in India. It is estimated that globally 58% of ADD deaths are attributed to inadequate drinking water, sanitation and hygiene [4]. The WHO/UNICEF Joint Monitoring Program for Water Supply, Sanitation and Hygiene (JMP) 2017 report revealed that 844 million people worldwide lack access to basic drinking-water service and 2.3 billion lack basic sanitation services, while 892 million still practiced open defecation [5]. The National Family Health Survey (NFHS-4, 2015–16) reported that in India only 52% of urban households and 18% of rural households have piped water supply, and the main source of water supply among rural households is bore-wells or tube-wells (51%). It has been estimated that 39% of households in India (54% among rural households) have no toilet facility and practicing open defecation [6]. The “indigenous†populations are socially, culturally and economically isolated and usually lack access to basic drinking-water and sanitation services. Therefore, they are vulnerable to ADD outbreaks and other emerging and re-emerging diseases [7]. The United Nations estimates that there are 370 million indigenous people existing across 90 countries of the world. They constitute 5% of the world population but 15% of the poorest [8]. India alone houses more than 705 such indigenous groups termed as Scheduled Tribes. As per the Census 2011, the total Scheduled Tribe population of India is 10.43 crore with a significant proportion of them living in rural areas [9]. On June 26, 2017, Kama-reddy district of Telangana state reported 55 ADD cases from the Pedda-GujjulThanda village. We conducted the outbreak investigation to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods Setting Pedda-Gujjul-Thanda village is a small tribal village with a total population of 563. The village is remotely located as an isolated community with a hilly terrain and is resource-limited with poor accessibility to sanitation and hygiene facilities. The nearest health care facility available for the residents is located at a distance of 10 km from the village. Case definition We defined a case as three or more loose stools within 24 h in a resident of the Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Case finding To find cases, we reviewed medical records of local health care facilities accessed by village residents in the nearby town. We conducted a medical camp in the village during the outbreak period for five days. We conducted a house-to-house survey in the village to find more cases, which are niether reported to health facility nor medical camp. Retrospective cohort study We conducted a retrospective cohort study to identify risk factors associated with illness. We defined the cohort as residents of Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Village resident was the unit of analysis. For data collection, we trained five teams of local paramedical staff. Using a pre-structured questionnaire, we collected data on demographic characteristics and risk factors related to drinking water, sanitation and hygiene. Good hand-washing practice was defined as reported washing of hands with soap and water every time after defecation and before eating. A bore-well less than 30-m-deep, as assessed from the records of village administration, was considered a shallow bore-well. Laboratory and environmental investigations Two stool samples were collected by the treating physician from admitted patients on the first day of hospital admission and transported to the state reference laboratory within two hours in Cary-Blair transport medium. The samples were cultured for Vibrio cholerae, Salmonella and Shigella on nutrient agar, MacConkey agar and deoxycholate citrate agar. Enteric pathogens were identified by biochemical reaction and by agglutination with anti-sera. We collected details of recent rainfall and conducted an environmental survey with household as sampling unit to assess drinking water, sanitation and hygiene practices. We assessed availability of residual chlorine in all village bore-wells and tested four of five bore-wells in the most affected colonies for faecal contamination by H2S method in field. Water was filled up to the ‘fill line’ of the sample bottle and incubated at room temperature (250–370 C) for 36–48 h and observed for colour change in the medium. A water sample was suspected to be contaminated with faecal matter, if it turned black [10, 11]. Because of limited supplies, we were unable to assess the fifth bore-well. Data analysis We analysed the data to describe the occurrence of cases over time, place, and person. We calculated relative risks (RR) with 95% confidence intervals (CI), population attributable risk percentages and conducted multiple logistic regression analysis with the dependent variables including consumption of shallow-downhill bore-well water, report of visible contaminants like mud in drinking water, illiteracy, household water treatment and good hand-washing practice. We used Epi Info version 7.2 for statistical analysis. Results Descriptive epidemiology We identified 191 ADD cases (65% females), with a village attack rate (AR) of 37% (191/512). The attack rate increased with age, with highest among > 60-year age group (55%) and lowest among children under-10 years (11%) (Table 1). No deaths were reported. In addition to diarrhea, cases presented with fever (17%), vomiting (16%) and abdominal pain (13%). 72% (138/191) cases reported to health care facilities and the medical camp conducted in the village. Among the 191 cases, 159 (83%) had mild illness treated with oral rehydration solution; 30 (16%) had moderate dehydration treated with intravenous fluids on out-patient basis, and 2 (1%) with severe dehydration were admitted in the district hospital for treatment with antibiotics (metronidazole and ciprofloxacin) and intravenous fluids. Cases started reported on June 26, 2017, with onset of symptoms from 24 June 2017. Maximum cases were reported on June 27, 2017, and no new cases were reported after June 30, 2017 (Fig. 1). The tribal population in the village had four sub-tribes namely Katroth, Badhawath, Nenawath, Baromath who resided in seven geographically demarcated colonies (labelled as A to G). Katroth sub-tribe resided in colonies A, B and G; Badhawath in colonies C and D; Nenawath in colony E and Baromath in colony F (Table 2). Colonies B and C had higher attack rates (65 and 47% respectively) as compared to other colonies (Fig. 2). Retrospective cohort study Among 563 village residents, 512 (91%) participated in the study. Among the 512 participants, median age was 28 years (range 1–80 years) with 52% females; 50% reported as illiterate with agriculture as the main source of livelihood for 76%. We analysed possible risk factors associated with ADD (Table 3). Drinking water from bore-well groundwater (vs canned water) was found significantly associated with ADD (RR = 12.7; 95% CI = 1.8–87.4). However, only 32 (6%) residents in the village used canned water and bore-well groundwater was the predominant source of water supply. Therefore, we analysed the water sources further, by location and type of bore-wells. Residents who used any of the five shallow bore-wells located downhill were significantly at higher risk (RR = 4.6; 95% CI = 3.4–6.1) and deep bore-wells were protective (RR = 0.4; 95% CI = 0.2–0.9). Report of visible contaminants like mud in drinking water (aOR = 4; 95% CI = 2.1–7.6) and illiteracy (aOR = 3.6; 95% CI = 3.5–10.1) were significantly associated with illness; and household water treatment (done either by boiling or use of candle filters) (aOR = 0.4; 95% CI = 0.2–0.7) and good hand-washing practice (aOR = 0.2; 95% CI = 0.1–0.5) were found protective. Laboratory and environmental results Stool samples collected from two hospitalized cases showed no growth for Vibrio cholerae, Salmonella and Shigella on culture. Among 110 households, 100 (91%) were available for environmental survey. Among the 100 houses surveyed, 79 (79%) were kutcha (low quality) type, made of mud, thatch and other low-quality material. Only 5 (5%) households had a designated toilet at home while the remaining 95 (95%) practiced open defecation at a site located on the slope of the hill behind the downhill colonies B and C (Figs. 2 and 3). Bore-wells were the main source of drinking water supply for 93 (93%) households. There were two deep borewells provided by the village administration and 17 shallow type bore-wells privately constructed by village residents. Five of these 17 (30%) shallow bore-wells were located in colonies B and C, on the downhill slope below the open defecation site. Plastic pipelines from the shallow wells were improperly installed with leakages at multiple points. There was no facility at source, for chlorination or any other mode of purification. Thirty households (30%) treated the water before consumption either by boiling or by use of candle filters. There was no routine drinking water surveillance in place by any authority for assessing the quality and fitness for drinking water. There was no residual chlorine found in any water samples. Three of four drinking water samples from bore-wells of most affected colonies (B and C) indicated faecal contamination by H2S field testing. There was heavy rainfall (average 65 mm in a day) from 22 to 24 June 2017. Prevention and control measures undertaken to contain the outbreak The village residents were discouraged from using shallow bore-well water and were provided with safe canned drinking water until all leakages were secured. Leakages in water supply from the bore-wells were identified and secured. Chlorine tablets were distributed for household level water disinfection. We informed the residents to avoid open defecation near drinking water sources and residential premises. Public health staff conducted health education daily to improve awareness among the villagers about water, sanitation, and hygiene. After active implementation of these control measures, cases declined rapidly in the village (Fig. 1). Discussion A rapid systematic epidemiological investigation of this outbreak identified water contamination points and likely mode of contamination. Based on these findings and our recommendations, the local health department instituted immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families. Effective implementation of public health measures limited the exposure of the community to contaminated water source resulting in rapid containment of the outbreak. Waterborne disease outbreaks tend to have cases spread over a time-period due to ongoing exposure to the contaminated water. In contrast, the pattern of epidemic curve in the present waterborne disease outbreak resembled that of food-borne with a point source exposure. Heavy rains contributed to the run-off of water from the open defecation site into the ground water of shallow wells located on slopes of hilly terrain resulting in heavy contamination and sudden rise of cases. Rapid control measures in the small village, implemented effectively within a short period of time, may have led to rapid decline of cases. The available epidemiological evidence also did not support generation of hypothesis of food-borne origin of the outbreak. In an outbreak reported among school children in Northern Greece in 2012, investigation revealed a waterborne viral gastroenteritis outbreak with a point source pattern, due to consumption of heavily contaminated water from a tap, which was not in use for two weeks during Christmas vacation [12]. Attack rate was high in this outbreak (37%), possibly due to exposure to high pathogen load subsequent to gross faecal contamination of water sources. In the absence of other alternative water sources, this tribal community was exclusively dependent on the contaminated water source for drinking, therefore exposing a large section of the community to risk. Geetha et al. analysed 32 diarrheal outbreaks in south India in non- tribal communities and reported lower attack rates varying from 0.6 to 21.5% [13]. However, tribal populations in India such as in Pedda-Gujjul-Thanda are marginalized with poor availability of WASH facilities [14]. This vulnerable tribal population continues to be at higher risk for ADD outbreaks with 27% not having access to safe drinking water and 75% of households not having toilets [15]. They need special assistance schemes from the government to enable them overcome poor accessibility to WASH facilities and secure healthy living [16]. Due to inadequate availability of communally managed safe public water points by the local authority, this community in Pedda-Gujjul-Thanda village was dependent on privately constructed shallow bore-wells for water supply. These are economical but likely to be unsafe. In this outbreak, open defecation site was present on the downhill slope in proximity to the residential premises and water resources, increasing the risk of drinking water contamination. Among the entire village population, 61% of ADD cases were attributable to drinking water from the ‘shallow downhill bore-wells’ (Population Attributable Fraction 61%), which was also evident from rapid outbreak containment following the elimination of exposure to this single risk factor. Since this exposure factor is amenable to long-term public health intervention, permanent elimination of shallow downhill borewells as water source was recommended, replacing them with properly secured deep bore-wells. Shallow bore-wells are known for their susceptibility to contamination from surface land-use activities [17, 18]. Studies have found levels of E. coli and enteric viruses to be high in shallow sources of ground water especially when they are in close proximity to polluting sources [19–21]. Consumption of ground water from shallow bore-wells with no purification facility increases the risk of diarrhea outbreaks manifold [22, 23]. A metaanalytic study of water-borne diarrheal disease outbreaks in China reported that 78 of 85 (92%) outbreaks (between year 1987 to 2014) were due to poor sanitary conditions of wells with lavatories/septic tanks nearby and lack of purification facilities [24]. In developed countries and urban areas of developing countries, as water supply and sanitation have improved dramatically over a period of time, such outbreaks were rarely reported in the recent past. The largest E. coli O157 outbreak in United States occurred in 1999 at a county fair (781 ill persons and 2 deaths) was due to groundwater source from a temporary unregulated well at the fairground [25]. Our findings have implication for India’s progress towards United Nation’s Sustainable Development Goal (SDG) 6 and India’s nation-wide campaign ‘Swachh Bharat Mission (SBM)’ to ensure availability and management of water and sanitation for all. SDG 6 aims at achieving universal access to basic sanitation service by 2030; and it has been reported that between 2000 and 2017, the proportion lacking even a basic sanitation service decreased from 44 to 27% [26]. SBM aims to achieve an open-defecation free status in rural areas through the construction of household-owned and community-owned toilets and establishing an accountable mechanism of monitoring toilet use. In 2015 in India, around 524 million (39%) practiced open defecation. However, under the SBM mission, due to increase in ‘households with toilets’ only 19 million (1.4%) practiced open defecation in January 2019 [5, 27]. There has also been a 71.58% increase in ‘households with toilets’ from October 2014 to October 2019 in rural areas of the Telangana state in India [27]. The tribal community initially obstructed the effective delivery of health care services; however, after involvement of the local stakeholders and tribal leaders, the acceptance towards medical treatment and community health services improved. Notwithstanding, most of the patients were still reluctant and did not consent for giving stool specimens for laboratory diagnosis. Establishing a rapport with the reticent tribal community was a major challenge faced by the outbreak investigation team. Lack of microbiological aetiology confirmation of the outbreak remained a limitation of the investigation due to limited stool samples and laboratory-capacity constraints of the remote area. Recognizing the pivotal importance of SDGs, national health policy of India (2017) has set the health-related cross-sectorial goal “access to safe water and sanitation to all by 2020†[28]. Greater political and financial commitment towards resource-limited remote tribal areas with effective community mobilization is required to accelerate the public health interventions to improve WASH and to prevent ADD outbreaks in the future. Conclusion This was a community-wide acute diarrheal disease outbreak with high village attack rate in a remote tribal village of Telangana with poor availability of safe water, sanitation and hygiene (WASH) facilities. A rapid and systematic epidemiological investigation identified drinking of faecal-contaminated water from the shallow borewells as the leading cause for this outbreak. These borewells were likely contaminated from runoff after rain from open defecation areas located on a downhill slope. Prompt and targeted public health action contained the number of cases.

Where did this happen?

92

A community-wide acute diarrheal disease outbreak associated with drinking contaminated water from shallow bore-wells in a tribal village

Abstract Background: In 2016, India reported 709 acute diarrheal disease (ADD) outbreaks (> 25% of all outbreaks). Tribal populations are at higher risk with 27% not having accessibility to safe drinking water and 75% households not having toilets. On June 26, 2017 Pedda-Gujjul-Thanda, a tribal village reported an acute diarrheal disease (ADD) outbreak. We investigated to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods: We defined a case as ≥3 loose stools within 24 h in Pedda-Gujjul-Thanda residents from June 24–30, 2017. We identified cases by reviewing hospital records and house-to-house survey. We conducted a retrospective cohort study and collected stool samples for culture. We assessed drinking water supply and sanitation practices and tested water samples for faecal-contamination. Results: We identified 191 cases (65% females) with median age 36 years (range 4–80 years) and no deaths. The attack-rate (AR) was 37% (191/512). Downhill colonies (located on slope of hilly terrains of the village) reported higher ARs (56%[136/243], p < 0.001) than others (20%[55/269]). Symptoms included diarrhea (100%), fever (17%), vomiting (16%) and abdominal pain (13%). Drinking water from five shallow bore-wells located in downhill colonies was significantly associated with illness (RR = 4.6, 95%CI = 3.4–6.1 and population attributable fraction 61%). In multivariate analysis, drinking water from the shallow bore-wells located in downhill colonies (aOR = 7.9, [95% CI =4.7– 13.2]), illiteracy (aOR =6, [95% CI = 3.6–10.1]), good hand-washing practice (aOR = 0.4, [95%CI = 0.2–0.7]) and household water treatment (aOR = 0.3, [95%CI = 0.2–0.5]) were significantly associated with illness. Two stool cultures were negative for Vibrio cholerae. Heavy rainfall was reported from June 22–24. Five of six water samples collected from shallow bore-wells located in downhill colonies were positive for faecal contamination. Conclusion: An ADD outbreak with high attack rate in a remote tribal village was associated with drinking water from shallow downhill bore-wells, likely contaminated via runoff from open defecation areas after heavy rains. Based on our recommendations, immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families, and as long-term public health measures construction of house-hold latrines and piped-water supply initiated. Keywords: Acute diarrheal disease, Outbreak, Bore-well, Tribal Introduction Globally there are an estimated 1.7 billion cases and 2.2 million deaths from acute diarrheal disease (ADD) every year [1]. In India, the burden is particularly high with more than 13.9 million cases reported in 2016 and 709 ADD outbreaks reported accounting to more than 25% of all outbreaks [2, 3]. Lack of access to safe drinking water and basic sanitation are the leading causes of ADD burden globally and in India. It is estimated that globally 58% of ADD deaths are attributed to inadequate drinking water, sanitation and hygiene [4]. The WHO/UNICEF Joint Monitoring Program for Water Supply, Sanitation and Hygiene (JMP) 2017 report revealed that 844 million people worldwide lack access to basic drinking-water service and 2.3 billion lack basic sanitation services, while 892 million still practiced open defecation [5]. The National Family Health Survey (NFHS-4, 2015–16) reported that in India only 52% of urban households and 18% of rural households have piped water supply, and the main source of water supply among rural households is bore-wells or tube-wells (51%). It has been estimated that 39% of households in India (54% among rural households) have no toilet facility and practicing open defecation [6]. The “indigenous†populations are socially, culturally and economically isolated and usually lack access to basic drinking-water and sanitation services. Therefore, they are vulnerable to ADD outbreaks and other emerging and re-emerging diseases [7]. The United Nations estimates that there are 370 million indigenous people existing across 90 countries of the world. They constitute 5% of the world population but 15% of the poorest [8]. India alone houses more than 705 such indigenous groups termed as Scheduled Tribes. As per the Census 2011, the total Scheduled Tribe population of India is 10.43 crore with a significant proportion of them living in rural areas [9]. On June 26, 2017, Kama-reddy district of Telangana state reported 55 ADD cases from the Pedda-GujjulThanda village. We conducted the outbreak investigation to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods Setting Pedda-Gujjul-Thanda village is a small tribal village with a total population of 563. The village is remotely located as an isolated community with a hilly terrain and is resource-limited with poor accessibility to sanitation and hygiene facilities. The nearest health care facility available for the residents is located at a distance of 10 km from the village. Case definition We defined a case as three or more loose stools within 24 h in a resident of the Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Case finding To find cases, we reviewed medical records of local health care facilities accessed by village residents in the nearby town. We conducted a medical camp in the village during the outbreak period for five days. We conducted a house-to-house survey in the village to find more cases, which are niether reported to health facility nor medical camp. Retrospective cohort study We conducted a retrospective cohort study to identify risk factors associated with illness. We defined the cohort as residents of Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Village resident was the unit of analysis. For data collection, we trained five teams of local paramedical staff. Using a pre-structured questionnaire, we collected data on demographic characteristics and risk factors related to drinking water, sanitation and hygiene. Good hand-washing practice was defined as reported washing of hands with soap and water every time after defecation and before eating. A bore-well less than 30-m-deep, as assessed from the records of village administration, was considered a shallow bore-well. Laboratory and environmental investigations Two stool samples were collected by the treating physician from admitted patients on the first day of hospital admission and transported to the state reference laboratory within two hours in Cary-Blair transport medium. The samples were cultured for Vibrio cholerae, Salmonella and Shigella on nutrient agar, MacConkey agar and deoxycholate citrate agar. Enteric pathogens were identified by biochemical reaction and by agglutination with anti-sera. We collected details of recent rainfall and conducted an environmental survey with household as sampling unit to assess drinking water, sanitation and hygiene practices. We assessed availability of residual chlorine in all village bore-wells and tested four of five bore-wells in the most affected colonies for faecal contamination by H2S method in field. Water was filled up to the ‘fill line’ of the sample bottle and incubated at room temperature (250–370 C) for 36–48 h and observed for colour change in the medium. A water sample was suspected to be contaminated with faecal matter, if it turned black [10, 11]. Because of limited supplies, we were unable to assess the fifth bore-well. Data analysis We analysed the data to describe the occurrence of cases over time, place, and person. We calculated relative risks (RR) with 95% confidence intervals (CI), population attributable risk percentages and conducted multiple logistic regression analysis with the dependent variables including consumption of shallow-downhill bore-well water, report of visible contaminants like mud in drinking water, illiteracy, household water treatment and good hand-washing practice. We used Epi Info version 7.2 for statistical analysis. Results Descriptive epidemiology We identified 191 ADD cases (65% females), with a village attack rate (AR) of 37% (191/512). The attack rate increased with age, with highest among > 60-year age group (55%) and lowest among children under-10 years (11%) (Table 1). No deaths were reported. In addition to diarrhea, cases presented with fever (17%), vomiting (16%) and abdominal pain (13%). 72% (138/191) cases reported to health care facilities and the medical camp conducted in the village. Among the 191 cases, 159 (83%) had mild illness treated with oral rehydration solution; 30 (16%) had moderate dehydration treated with intravenous fluids on out-patient basis, and 2 (1%) with severe dehydration were admitted in the district hospital for treatment with antibiotics (metronidazole and ciprofloxacin) and intravenous fluids. Cases started reported on June 26, 2017, with onset of symptoms from 24 June 2017. Maximum cases were reported on June 27, 2017, and no new cases were reported after June 30, 2017 (Fig. 1). The tribal population in the village had four sub-tribes namely Katroth, Badhawath, Nenawath, Baromath who resided in seven geographically demarcated colonies (labelled as A to G). Katroth sub-tribe resided in colonies A, B and G; Badhawath in colonies C and D; Nenawath in colony E and Baromath in colony F (Table 2). Colonies B and C had higher attack rates (65 and 47% respectively) as compared to other colonies (Fig. 2). Retrospective cohort study Among 563 village residents, 512 (91%) participated in the study. Among the 512 participants, median age was 28 years (range 1–80 years) with 52% females; 50% reported as illiterate with agriculture as the main source of livelihood for 76%. We analysed possible risk factors associated with ADD (Table 3). Drinking water from bore-well groundwater (vs canned water) was found significantly associated with ADD (RR = 12.7; 95% CI = 1.8–87.4). However, only 32 (6%) residents in the village used canned water and bore-well groundwater was the predominant source of water supply. Therefore, we analysed the water sources further, by location and type of bore-wells. Residents who used any of the five shallow bore-wells located downhill were significantly at higher risk (RR = 4.6; 95% CI = 3.4–6.1) and deep bore-wells were protective (RR = 0.4; 95% CI = 0.2–0.9). Report of visible contaminants like mud in drinking water (aOR = 4; 95% CI = 2.1–7.6) and illiteracy (aOR = 3.6; 95% CI = 3.5–10.1) were significantly associated with illness; and household water treatment (done either by boiling or use of candle filters) (aOR = 0.4; 95% CI = 0.2–0.7) and good hand-washing practice (aOR = 0.2; 95% CI = 0.1–0.5) were found protective. Laboratory and environmental results Stool samples collected from two hospitalized cases showed no growth for Vibrio cholerae, Salmonella and Shigella on culture. Among 110 households, 100 (91%) were available for environmental survey. Among the 100 houses surveyed, 79 (79%) were kutcha (low quality) type, made of mud, thatch and other low-quality material. Only 5 (5%) households had a designated toilet at home while the remaining 95 (95%) practiced open defecation at a site located on the slope of the hill behind the downhill colonies B and C (Figs. 2 and 3). Bore-wells were the main source of drinking water supply for 93 (93%) households. There were two deep borewells provided by the village administration and 17 shallow type bore-wells privately constructed by village residents. Five of these 17 (30%) shallow bore-wells were located in colonies B and C, on the downhill slope below the open defecation site. Plastic pipelines from the shallow wells were improperly installed with leakages at multiple points. There was no facility at source, for chlorination or any other mode of purification. Thirty households (30%) treated the water before consumption either by boiling or by use of candle filters. There was no routine drinking water surveillance in place by any authority for assessing the quality and fitness for drinking water. There was no residual chlorine found in any water samples. Three of four drinking water samples from bore-wells of most affected colonies (B and C) indicated faecal contamination by H2S field testing. There was heavy rainfall (average 65 mm in a day) from 22 to 24 June 2017. Prevention and control measures undertaken to contain the outbreak The village residents were discouraged from using shallow bore-well water and were provided with safe canned drinking water until all leakages were secured. Leakages in water supply from the bore-wells were identified and secured. Chlorine tablets were distributed for household level water disinfection. We informed the residents to avoid open defecation near drinking water sources and residential premises. Public health staff conducted health education daily to improve awareness among the villagers about water, sanitation, and hygiene. After active implementation of these control measures, cases declined rapidly in the village (Fig. 1). Discussion A rapid systematic epidemiological investigation of this outbreak identified water contamination points and likely mode of contamination. Based on these findings and our recommendations, the local health department instituted immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families. Effective implementation of public health measures limited the exposure of the community to contaminated water source resulting in rapid containment of the outbreak. Waterborne disease outbreaks tend to have cases spread over a time-period due to ongoing exposure to the contaminated water. In contrast, the pattern of epidemic curve in the present waterborne disease outbreak resembled that of food-borne with a point source exposure. Heavy rains contributed to the run-off of water from the open defecation site into the ground water of shallow wells located on slopes of hilly terrain resulting in heavy contamination and sudden rise of cases. Rapid control measures in the small village, implemented effectively within a short period of time, may have led to rapid decline of cases. The available epidemiological evidence also did not support generation of hypothesis of food-borne origin of the outbreak. In an outbreak reported among school children in Northern Greece in 2012, investigation revealed a waterborne viral gastroenteritis outbreak with a point source pattern, due to consumption of heavily contaminated water from a tap, which was not in use for two weeks during Christmas vacation [12]. Attack rate was high in this outbreak (37%), possibly due to exposure to high pathogen load subsequent to gross faecal contamination of water sources. In the absence of other alternative water sources, this tribal community was exclusively dependent on the contaminated water source for drinking, therefore exposing a large section of the community to risk. Geetha et al. analysed 32 diarrheal outbreaks in south India in non- tribal communities and reported lower attack rates varying from 0.6 to 21.5% [13]. However, tribal populations in India such as in Pedda-Gujjul-Thanda are marginalized with poor availability of WASH facilities [14]. This vulnerable tribal population continues to be at higher risk for ADD outbreaks with 27% not having access to safe drinking water and 75% of households not having toilets [15]. They need special assistance schemes from the government to enable them overcome poor accessibility to WASH facilities and secure healthy living [16]. Due to inadequate availability of communally managed safe public water points by the local authority, this community in Pedda-Gujjul-Thanda village was dependent on privately constructed shallow bore-wells for water supply. These are economical but likely to be unsafe. In this outbreak, open defecation site was present on the downhill slope in proximity to the residential premises and water resources, increasing the risk of drinking water contamination. Among the entire village population, 61% of ADD cases were attributable to drinking water from the ‘shallow downhill bore-wells’ (Population Attributable Fraction 61%), which was also evident from rapid outbreak containment following the elimination of exposure to this single risk factor. Since this exposure factor is amenable to long-term public health intervention, permanent elimination of shallow downhill borewells as water source was recommended, replacing them with properly secured deep bore-wells. Shallow bore-wells are known for their susceptibility to contamination from surface land-use activities [17, 18]. Studies have found levels of E. coli and enteric viruses to be high in shallow sources of ground water especially when they are in close proximity to polluting sources [19–21]. Consumption of ground water from shallow bore-wells with no purification facility increases the risk of diarrhea outbreaks manifold [22, 23]. A metaanalytic study of water-borne diarrheal disease outbreaks in China reported that 78 of 85 (92%) outbreaks (between year 1987 to 2014) were due to poor sanitary conditions of wells with lavatories/septic tanks nearby and lack of purification facilities [24]. In developed countries and urban areas of developing countries, as water supply and sanitation have improved dramatically over a period of time, such outbreaks were rarely reported in the recent past. The largest E. coli O157 outbreak in United States occurred in 1999 at a county fair (781 ill persons and 2 deaths) was due to groundwater source from a temporary unregulated well at the fairground [25]. Our findings have implication for India’s progress towards United Nation’s Sustainable Development Goal (SDG) 6 and India’s nation-wide campaign ‘Swachh Bharat Mission (SBM)’ to ensure availability and management of water and sanitation for all. SDG 6 aims at achieving universal access to basic sanitation service by 2030; and it has been reported that between 2000 and 2017, the proportion lacking even a basic sanitation service decreased from 44 to 27% [26]. SBM aims to achieve an open-defecation free status in rural areas through the construction of household-owned and community-owned toilets and establishing an accountable mechanism of monitoring toilet use. In 2015 in India, around 524 million (39%) practiced open defecation. However, under the SBM mission, due to increase in ‘households with toilets’ only 19 million (1.4%) practiced open defecation in January 2019 [5, 27]. There has also been a 71.58% increase in ‘households with toilets’ from October 2014 to October 2019 in rural areas of the Telangana state in India [27]. The tribal community initially obstructed the effective delivery of health care services; however, after involvement of the local stakeholders and tribal leaders, the acceptance towards medical treatment and community health services improved. Notwithstanding, most of the patients were still reluctant and did not consent for giving stool specimens for laboratory diagnosis. Establishing a rapport with the reticent tribal community was a major challenge faced by the outbreak investigation team. Lack of microbiological aetiology confirmation of the outbreak remained a limitation of the investigation due to limited stool samples and laboratory-capacity constraints of the remote area. Recognizing the pivotal importance of SDGs, national health policy of India (2017) has set the health-related cross-sectorial goal “access to safe water and sanitation to all by 2020†[28]. Greater political and financial commitment towards resource-limited remote tribal areas with effective community mobilization is required to accelerate the public health interventions to improve WASH and to prevent ADD outbreaks in the future. Conclusion This was a community-wide acute diarrheal disease outbreak with high village attack rate in a remote tribal village of Telangana with poor availability of safe water, sanitation and hygiene (WASH) facilities. A rapid and systematic epidemiological investigation identified drinking of faecal-contaminated water from the shallow borewells as the leading cause for this outbreak. These borewells were likely contaminated from runoff after rain from open defecation areas located on a downhill slope. Prompt and targeted public health action contained the number of cases.

What caused the event?

93

A community-wide acute diarrheal disease outbreak associated with drinking contaminated water from shallow bore-wells in a tribal village

Abstract Background: In 2016, India reported 709 acute diarrheal disease (ADD) outbreaks (> 25% of all outbreaks). Tribal populations are at higher risk with 27% not having accessibility to safe drinking water and 75% households not having toilets. On June 26, 2017 Pedda-Gujjul-Thanda, a tribal village reported an acute diarrheal disease (ADD) outbreak. We investigated to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods: We defined a case as ≥3 loose stools within 24 h in Pedda-Gujjul-Thanda residents from June 24–30, 2017. We identified cases by reviewing hospital records and house-to-house survey. We conducted a retrospective cohort study and collected stool samples for culture. We assessed drinking water supply and sanitation practices and tested water samples for faecal-contamination. Results: We identified 191 cases (65% females) with median age 36 years (range 4–80 years) and no deaths. The attack-rate (AR) was 37% (191/512). Downhill colonies (located on slope of hilly terrains of the village) reported higher ARs (56%[136/243], p < 0.001) than others (20%[55/269]). Symptoms included diarrhea (100%), fever (17%), vomiting (16%) and abdominal pain (13%). Drinking water from five shallow bore-wells located in downhill colonies was significantly associated with illness (RR = 4.6, 95%CI = 3.4–6.1 and population attributable fraction 61%). In multivariate analysis, drinking water from the shallow bore-wells located in downhill colonies (aOR = 7.9, [95% CI =4.7– 13.2]), illiteracy (aOR =6, [95% CI = 3.6–10.1]), good hand-washing practice (aOR = 0.4, [95%CI = 0.2–0.7]) and household water treatment (aOR = 0.3, [95%CI = 0.2–0.5]) were significantly associated with illness. Two stool cultures were negative for Vibrio cholerae. Heavy rainfall was reported from June 22–24. Five of six water samples collected from shallow bore-wells located in downhill colonies were positive for faecal contamination. Conclusion: An ADD outbreak with high attack rate in a remote tribal village was associated with drinking water from shallow downhill bore-wells, likely contaminated via runoff from open defecation areas after heavy rains. Based on our recommendations, immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families, and as long-term public health measures construction of house-hold latrines and piped-water supply initiated. Keywords: Acute diarrheal disease, Outbreak, Bore-well, Tribal Introduction Globally there are an estimated 1.7 billion cases and 2.2 million deaths from acute diarrheal disease (ADD) every year [1]. In India, the burden is particularly high with more than 13.9 million cases reported in 2016 and 709 ADD outbreaks reported accounting to more than 25% of all outbreaks [2, 3]. Lack of access to safe drinking water and basic sanitation are the leading causes of ADD burden globally and in India. It is estimated that globally 58% of ADD deaths are attributed to inadequate drinking water, sanitation and hygiene [4]. The WHO/UNICEF Joint Monitoring Program for Water Supply, Sanitation and Hygiene (JMP) 2017 report revealed that 844 million people worldwide lack access to basic drinking-water service and 2.3 billion lack basic sanitation services, while 892 million still practiced open defecation [5]. The National Family Health Survey (NFHS-4, 2015–16) reported that in India only 52% of urban households and 18% of rural households have piped water supply, and the main source of water supply among rural households is bore-wells or tube-wells (51%). It has been estimated that 39% of households in India (54% among rural households) have no toilet facility and practicing open defecation [6]. The “indigenous†populations are socially, culturally and economically isolated and usually lack access to basic drinking-water and sanitation services. Therefore, they are vulnerable to ADD outbreaks and other emerging and re-emerging diseases [7]. The United Nations estimates that there are 370 million indigenous people existing across 90 countries of the world. They constitute 5% of the world population but 15% of the poorest [8]. India alone houses more than 705 such indigenous groups termed as Scheduled Tribes. As per the Census 2011, the total Scheduled Tribe population of India is 10.43 crore with a significant proportion of them living in rural areas [9]. On June 26, 2017, Kama-reddy district of Telangana state reported 55 ADD cases from the Pedda-GujjulThanda village. We conducted the outbreak investigation to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods Setting Pedda-Gujjul-Thanda village is a small tribal village with a total population of 563. The village is remotely located as an isolated community with a hilly terrain and is resource-limited with poor accessibility to sanitation and hygiene facilities. The nearest health care facility available for the residents is located at a distance of 10 km from the village. Case definition We defined a case as three or more loose stools within 24 h in a resident of the Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Case finding To find cases, we reviewed medical records of local health care facilities accessed by village residents in the nearby town. We conducted a medical camp in the village during the outbreak period for five days. We conducted a house-to-house survey in the village to find more cases, which are niether reported to health facility nor medical camp. Retrospective cohort study We conducted a retrospective cohort study to identify risk factors associated with illness. We defined the cohort as residents of Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Village resident was the unit of analysis. For data collection, we trained five teams of local paramedical staff. Using a pre-structured questionnaire, we collected data on demographic characteristics and risk factors related to drinking water, sanitation and hygiene. Good hand-washing practice was defined as reported washing of hands with soap and water every time after defecation and before eating. A bore-well less than 30-m-deep, as assessed from the records of village administration, was considered a shallow bore-well. Laboratory and environmental investigations Two stool samples were collected by the treating physician from admitted patients on the first day of hospital admission and transported to the state reference laboratory within two hours in Cary-Blair transport medium. The samples were cultured for Vibrio cholerae, Salmonella and Shigella on nutrient agar, MacConkey agar and deoxycholate citrate agar. Enteric pathogens were identified by biochemical reaction and by agglutination with anti-sera. We collected details of recent rainfall and conducted an environmental survey with household as sampling unit to assess drinking water, sanitation and hygiene practices. We assessed availability of residual chlorine in all village bore-wells and tested four of five bore-wells in the most affected colonies for faecal contamination by H2S method in field. Water was filled up to the ‘fill line’ of the sample bottle and incubated at room temperature (250–370 C) for 36–48 h and observed for colour change in the medium. A water sample was suspected to be contaminated with faecal matter, if it turned black [10, 11]. Because of limited supplies, we were unable to assess the fifth bore-well. Data analysis We analysed the data to describe the occurrence of cases over time, place, and person. We calculated relative risks (RR) with 95% confidence intervals (CI), population attributable risk percentages and conducted multiple logistic regression analysis with the dependent variables including consumption of shallow-downhill bore-well water, report of visible contaminants like mud in drinking water, illiteracy, household water treatment and good hand-washing practice. We used Epi Info version 7.2 for statistical analysis. Results Descriptive epidemiology We identified 191 ADD cases (65% females), with a village attack rate (AR) of 37% (191/512). The attack rate increased with age, with highest among > 60-year age group (55%) and lowest among children under-10 years (11%) (Table 1). No deaths were reported. In addition to diarrhea, cases presented with fever (17%), vomiting (16%) and abdominal pain (13%). 72% (138/191) cases reported to health care facilities and the medical camp conducted in the village. Among the 191 cases, 159 (83%) had mild illness treated with oral rehydration solution; 30 (16%) had moderate dehydration treated with intravenous fluids on out-patient basis, and 2 (1%) with severe dehydration were admitted in the district hospital for treatment with antibiotics (metronidazole and ciprofloxacin) and intravenous fluids. Cases started reported on June 26, 2017, with onset of symptoms from 24 June 2017. Maximum cases were reported on June 27, 2017, and no new cases were reported after June 30, 2017 (Fig. 1). The tribal population in the village had four sub-tribes namely Katroth, Badhawath, Nenawath, Baromath who resided in seven geographically demarcated colonies (labelled as A to G). Katroth sub-tribe resided in colonies A, B and G; Badhawath in colonies C and D; Nenawath in colony E and Baromath in colony F (Table 2). Colonies B and C had higher attack rates (65 and 47% respectively) as compared to other colonies (Fig. 2). Retrospective cohort study Among 563 village residents, 512 (91%) participated in the study. Among the 512 participants, median age was 28 years (range 1–80 years) with 52% females; 50% reported as illiterate with agriculture as the main source of livelihood for 76%. We analysed possible risk factors associated with ADD (Table 3). Drinking water from bore-well groundwater (vs canned water) was found significantly associated with ADD (RR = 12.7; 95% CI = 1.8–87.4). However, only 32 (6%) residents in the village used canned water and bore-well groundwater was the predominant source of water supply. Therefore, we analysed the water sources further, by location and type of bore-wells. Residents who used any of the five shallow bore-wells located downhill were significantly at higher risk (RR = 4.6; 95% CI = 3.4–6.1) and deep bore-wells were protective (RR = 0.4; 95% CI = 0.2–0.9). Report of visible contaminants like mud in drinking water (aOR = 4; 95% CI = 2.1–7.6) and illiteracy (aOR = 3.6; 95% CI = 3.5–10.1) were significantly associated with illness; and household water treatment (done either by boiling or use of candle filters) (aOR = 0.4; 95% CI = 0.2–0.7) and good hand-washing practice (aOR = 0.2; 95% CI = 0.1–0.5) were found protective. Laboratory and environmental results Stool samples collected from two hospitalized cases showed no growth for Vibrio cholerae, Salmonella and Shigella on culture. Among 110 households, 100 (91%) were available for environmental survey. Among the 100 houses surveyed, 79 (79%) were kutcha (low quality) type, made of mud, thatch and other low-quality material. Only 5 (5%) households had a designated toilet at home while the remaining 95 (95%) practiced open defecation at a site located on the slope of the hill behind the downhill colonies B and C (Figs. 2 and 3). Bore-wells were the main source of drinking water supply for 93 (93%) households. There were two deep borewells provided by the village administration and 17 shallow type bore-wells privately constructed by village residents. Five of these 17 (30%) shallow bore-wells were located in colonies B and C, on the downhill slope below the open defecation site. Plastic pipelines from the shallow wells were improperly installed with leakages at multiple points. There was no facility at source, for chlorination or any other mode of purification. Thirty households (30%) treated the water before consumption either by boiling or by use of candle filters. There was no routine drinking water surveillance in place by any authority for assessing the quality and fitness for drinking water. There was no residual chlorine found in any water samples. Three of four drinking water samples from bore-wells of most affected colonies (B and C) indicated faecal contamination by H2S field testing. There was heavy rainfall (average 65 mm in a day) from 22 to 24 June 2017. Prevention and control measures undertaken to contain the outbreak The village residents were discouraged from using shallow bore-well water and were provided with safe canned drinking water until all leakages were secured. Leakages in water supply from the bore-wells were identified and secured. Chlorine tablets were distributed for household level water disinfection. We informed the residents to avoid open defecation near drinking water sources and residential premises. Public health staff conducted health education daily to improve awareness among the villagers about water, sanitation, and hygiene. After active implementation of these control measures, cases declined rapidly in the village (Fig. 1). Discussion A rapid systematic epidemiological investigation of this outbreak identified water contamination points and likely mode of contamination. Based on these findings and our recommendations, the local health department instituted immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families. Effective implementation of public health measures limited the exposure of the community to contaminated water source resulting in rapid containment of the outbreak. Waterborne disease outbreaks tend to have cases spread over a time-period due to ongoing exposure to the contaminated water. In contrast, the pattern of epidemic curve in the present waterborne disease outbreak resembled that of food-borne with a point source exposure. Heavy rains contributed to the run-off of water from the open defecation site into the ground water of shallow wells located on slopes of hilly terrain resulting in heavy contamination and sudden rise of cases. Rapid control measures in the small village, implemented effectively within a short period of time, may have led to rapid decline of cases. The available epidemiological evidence also did not support generation of hypothesis of food-borne origin of the outbreak. In an outbreak reported among school children in Northern Greece in 2012, investigation revealed a waterborne viral gastroenteritis outbreak with a point source pattern, due to consumption of heavily contaminated water from a tap, which was not in use for two weeks during Christmas vacation [12]. Attack rate was high in this outbreak (37%), possibly due to exposure to high pathogen load subsequent to gross faecal contamination of water sources. In the absence of other alternative water sources, this tribal community was exclusively dependent on the contaminated water source for drinking, therefore exposing a large section of the community to risk. Geetha et al. analysed 32 diarrheal outbreaks in south India in non- tribal communities and reported lower attack rates varying from 0.6 to 21.5% [13]. However, tribal populations in India such as in Pedda-Gujjul-Thanda are marginalized with poor availability of WASH facilities [14]. This vulnerable tribal population continues to be at higher risk for ADD outbreaks with 27% not having access to safe drinking water and 75% of households not having toilets [15]. They need special assistance schemes from the government to enable them overcome poor accessibility to WASH facilities and secure healthy living [16]. Due to inadequate availability of communally managed safe public water points by the local authority, this community in Pedda-Gujjul-Thanda village was dependent on privately constructed shallow bore-wells for water supply. These are economical but likely to be unsafe. In this outbreak, open defecation site was present on the downhill slope in proximity to the residential premises and water resources, increasing the risk of drinking water contamination. Among the entire village population, 61% of ADD cases were attributable to drinking water from the ‘shallow downhill bore-wells’ (Population Attributable Fraction 61%), which was also evident from rapid outbreak containment following the elimination of exposure to this single risk factor. Since this exposure factor is amenable to long-term public health intervention, permanent elimination of shallow downhill borewells as water source was recommended, replacing them with properly secured deep bore-wells. Shallow bore-wells are known for their susceptibility to contamination from surface land-use activities [17, 18]. Studies have found levels of E. coli and enteric viruses to be high in shallow sources of ground water especially when they are in close proximity to polluting sources [19–21]. Consumption of ground water from shallow bore-wells with no purification facility increases the risk of diarrhea outbreaks manifold [22, 23]. A metaanalytic study of water-borne diarrheal disease outbreaks in China reported that 78 of 85 (92%) outbreaks (between year 1987 to 2014) were due to poor sanitary conditions of wells with lavatories/septic tanks nearby and lack of purification facilities [24]. In developed countries and urban areas of developing countries, as water supply and sanitation have improved dramatically over a period of time, such outbreaks were rarely reported in the recent past. The largest E. coli O157 outbreak in United States occurred in 1999 at a county fair (781 ill persons and 2 deaths) was due to groundwater source from a temporary unregulated well at the fairground [25]. Our findings have implication for India’s progress towards United Nation’s Sustainable Development Goal (SDG) 6 and India’s nation-wide campaign ‘Swachh Bharat Mission (SBM)’ to ensure availability and management of water and sanitation for all. SDG 6 aims at achieving universal access to basic sanitation service by 2030; and it has been reported that between 2000 and 2017, the proportion lacking even a basic sanitation service decreased from 44 to 27% [26]. SBM aims to achieve an open-defecation free status in rural areas through the construction of household-owned and community-owned toilets and establishing an accountable mechanism of monitoring toilet use. In 2015 in India, around 524 million (39%) practiced open defecation. However, under the SBM mission, due to increase in ‘households with toilets’ only 19 million (1.4%) practiced open defecation in January 2019 [5, 27]. There has also been a 71.58% increase in ‘households with toilets’ from October 2014 to October 2019 in rural areas of the Telangana state in India [27]. The tribal community initially obstructed the effective delivery of health care services; however, after involvement of the local stakeholders and tribal leaders, the acceptance towards medical treatment and community health services improved. Notwithstanding, most of the patients were still reluctant and did not consent for giving stool specimens for laboratory diagnosis. Establishing a rapport with the reticent tribal community was a major challenge faced by the outbreak investigation team. Lack of microbiological aetiology confirmation of the outbreak remained a limitation of the investigation due to limited stool samples and laboratory-capacity constraints of the remote area. Recognizing the pivotal importance of SDGs, national health policy of India (2017) has set the health-related cross-sectorial goal “access to safe water and sanitation to all by 2020†[28]. Greater political and financial commitment towards resource-limited remote tribal areas with effective community mobilization is required to accelerate the public health interventions to improve WASH and to prevent ADD outbreaks in the future. Conclusion This was a community-wide acute diarrheal disease outbreak with high village attack rate in a remote tribal village of Telangana with poor availability of safe water, sanitation and hygiene (WASH) facilities. A rapid and systematic epidemiological investigation identified drinking of faecal-contaminated water from the shallow borewells as the leading cause for this outbreak. These borewells were likely contaminated from runoff after rain from open defecation areas located on a downhill slope. Prompt and targeted public health action contained the number of cases.

What caused the faecal-contaminated water from the shallow borewells?

94

A community-wide acute diarrheal disease outbreak associated with drinking contaminated water from shallow bore-wells in a tribal village

Abstract Background: In 2016, India reported 709 acute diarrheal disease (ADD) outbreaks (> 25% of all outbreaks). Tribal populations are at higher risk with 27% not having accessibility to safe drinking water and 75% households not having toilets. On June 26, 2017 Pedda-Gujjul-Thanda, a tribal village reported an acute diarrheal disease (ADD) outbreak. We investigated to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods: We defined a case as ≥3 loose stools within 24 h in Pedda-Gujjul-Thanda residents from June 24–30, 2017. We identified cases by reviewing hospital records and house-to-house survey. We conducted a retrospective cohort study and collected stool samples for culture. We assessed drinking water supply and sanitation practices and tested water samples for faecal-contamination. Results: We identified 191 cases (65% females) with median age 36 years (range 4–80 years) and no deaths. The attack-rate (AR) was 37% (191/512). Downhill colonies (located on slope of hilly terrains of the village) reported higher ARs (56%[136/243], p < 0.001) than others (20%[55/269]). Symptoms included diarrhea (100%), fever (17%), vomiting (16%) and abdominal pain (13%). Drinking water from five shallow bore-wells located in downhill colonies was significantly associated with illness (RR = 4.6, 95%CI = 3.4–6.1 and population attributable fraction 61%). In multivariate analysis, drinking water from the shallow bore-wells located in downhill colonies (aOR = 7.9, [95% CI =4.7– 13.2]), illiteracy (aOR =6, [95% CI = 3.6–10.1]), good hand-washing practice (aOR = 0.4, [95%CI = 0.2–0.7]) and household water treatment (aOR = 0.3, [95%CI = 0.2–0.5]) were significantly associated with illness. Two stool cultures were negative for Vibrio cholerae. Heavy rainfall was reported from June 22–24. Five of six water samples collected from shallow bore-wells located in downhill colonies were positive for faecal contamination. Conclusion: An ADD outbreak with high attack rate in a remote tribal village was associated with drinking water from shallow downhill bore-wells, likely contaminated via runoff from open defecation areas after heavy rains. Based on our recommendations, immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families, and as long-term public health measures construction of house-hold latrines and piped-water supply initiated. Keywords: Acute diarrheal disease, Outbreak, Bore-well, Tribal Introduction Globally there are an estimated 1.7 billion cases and 2.2 million deaths from acute diarrheal disease (ADD) every year [1]. In India, the burden is particularly high with more than 13.9 million cases reported in 2016 and 709 ADD outbreaks reported accounting to more than 25% of all outbreaks [2, 3]. Lack of access to safe drinking water and basic sanitation are the leading causes of ADD burden globally and in India. It is estimated that globally 58% of ADD deaths are attributed to inadequate drinking water, sanitation and hygiene [4]. The WHO/UNICEF Joint Monitoring Program for Water Supply, Sanitation and Hygiene (JMP) 2017 report revealed that 844 million people worldwide lack access to basic drinking-water service and 2.3 billion lack basic sanitation services, while 892 million still practiced open defecation [5]. The National Family Health Survey (NFHS-4, 2015–16) reported that in India only 52% of urban households and 18% of rural households have piped water supply, and the main source of water supply among rural households is bore-wells or tube-wells (51%). It has been estimated that 39% of households in India (54% among rural households) have no toilet facility and practicing open defecation [6]. The “indigenous†populations are socially, culturally and economically isolated and usually lack access to basic drinking-water and sanitation services. Therefore, they are vulnerable to ADD outbreaks and other emerging and re-emerging diseases [7]. The United Nations estimates that there are 370 million indigenous people existing across 90 countries of the world. They constitute 5% of the world population but 15% of the poorest [8]. India alone houses more than 705 such indigenous groups termed as Scheduled Tribes. As per the Census 2011, the total Scheduled Tribe population of India is 10.43 crore with a significant proportion of them living in rural areas [9]. On June 26, 2017, Kama-reddy district of Telangana state reported 55 ADD cases from the Pedda-GujjulThanda village. We conducted the outbreak investigation to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods Setting Pedda-Gujjul-Thanda village is a small tribal village with a total population of 563. The village is remotely located as an isolated community with a hilly terrain and is resource-limited with poor accessibility to sanitation and hygiene facilities. The nearest health care facility available for the residents is located at a distance of 10 km from the village. Case definition We defined a case as three or more loose stools within 24 h in a resident of the Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Case finding To find cases, we reviewed medical records of local health care facilities accessed by village residents in the nearby town. We conducted a medical camp in the village during the outbreak period for five days. We conducted a house-to-house survey in the village to find more cases, which are niether reported to health facility nor medical camp. Retrospective cohort study We conducted a retrospective cohort study to identify risk factors associated with illness. We defined the cohort as residents of Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Village resident was the unit of analysis. For data collection, we trained five teams of local paramedical staff. Using a pre-structured questionnaire, we collected data on demographic characteristics and risk factors related to drinking water, sanitation and hygiene. Good hand-washing practice was defined as reported washing of hands with soap and water every time after defecation and before eating. A bore-well less than 30-m-deep, as assessed from the records of village administration, was considered a shallow bore-well. Laboratory and environmental investigations Two stool samples were collected by the treating physician from admitted patients on the first day of hospital admission and transported to the state reference laboratory within two hours in Cary-Blair transport medium. The samples were cultured for Vibrio cholerae, Salmonella and Shigella on nutrient agar, MacConkey agar and deoxycholate citrate agar. Enteric pathogens were identified by biochemical reaction and by agglutination with anti-sera. We collected details of recent rainfall and conducted an environmental survey with household as sampling unit to assess drinking water, sanitation and hygiene practices. We assessed availability of residual chlorine in all village bore-wells and tested four of five bore-wells in the most affected colonies for faecal contamination by H2S method in field. Water was filled up to the ‘fill line’ of the sample bottle and incubated at room temperature (250–370 C) for 36–48 h and observed for colour change in the medium. A water sample was suspected to be contaminated with faecal matter, if it turned black [10, 11]. Because of limited supplies, we were unable to assess the fifth bore-well. Data analysis We analysed the data to describe the occurrence of cases over time, place, and person. We calculated relative risks (RR) with 95% confidence intervals (CI), population attributable risk percentages and conducted multiple logistic regression analysis with the dependent variables including consumption of shallow-downhill bore-well water, report of visible contaminants like mud in drinking water, illiteracy, household water treatment and good hand-washing practice. We used Epi Info version 7.2 for statistical analysis. Results Descriptive epidemiology We identified 191 ADD cases (65% females), with a village attack rate (AR) of 37% (191/512). The attack rate increased with age, with highest among > 60-year age group (55%) and lowest among children under-10 years (11%) (Table 1). No deaths were reported. In addition to diarrhea, cases presented with fever (17%), vomiting (16%) and abdominal pain (13%). 72% (138/191) cases reported to health care facilities and the medical camp conducted in the village. Among the 191 cases, 159 (83%) had mild illness treated with oral rehydration solution; 30 (16%) had moderate dehydration treated with intravenous fluids on out-patient basis, and 2 (1%) with severe dehydration were admitted in the district hospital for treatment with antibiotics (metronidazole and ciprofloxacin) and intravenous fluids. Cases started reported on June 26, 2017, with onset of symptoms from 24 June 2017. Maximum cases were reported on June 27, 2017, and no new cases were reported after June 30, 2017 (Fig. 1). The tribal population in the village had four sub-tribes namely Katroth, Badhawath, Nenawath, Baromath who resided in seven geographically demarcated colonies (labelled as A to G). Katroth sub-tribe resided in colonies A, B and G; Badhawath in colonies C and D; Nenawath in colony E and Baromath in colony F (Table 2). Colonies B and C had higher attack rates (65 and 47% respectively) as compared to other colonies (Fig. 2). Retrospective cohort study Among 563 village residents, 512 (91%) participated in the study. Among the 512 participants, median age was 28 years (range 1–80 years) with 52% females; 50% reported as illiterate with agriculture as the main source of livelihood for 76%. We analysed possible risk factors associated with ADD (Table 3). Drinking water from bore-well groundwater (vs canned water) was found significantly associated with ADD (RR = 12.7; 95% CI = 1.8–87.4). However, only 32 (6%) residents in the village used canned water and bore-well groundwater was the predominant source of water supply. Therefore, we analysed the water sources further, by location and type of bore-wells. Residents who used any of the five shallow bore-wells located downhill were significantly at higher risk (RR = 4.6; 95% CI = 3.4–6.1) and deep bore-wells were protective (RR = 0.4; 95% CI = 0.2–0.9). Report of visible contaminants like mud in drinking water (aOR = 4; 95% CI = 2.1–7.6) and illiteracy (aOR = 3.6; 95% CI = 3.5–10.1) were significantly associated with illness; and household water treatment (done either by boiling or use of candle filters) (aOR = 0.4; 95% CI = 0.2–0.7) and good hand-washing practice (aOR = 0.2; 95% CI = 0.1–0.5) were found protective. Laboratory and environmental results Stool samples collected from two hospitalized cases showed no growth for Vibrio cholerae, Salmonella and Shigella on culture. Among 110 households, 100 (91%) were available for environmental survey. Among the 100 houses surveyed, 79 (79%) were kutcha (low quality) type, made of mud, thatch and other low-quality material. Only 5 (5%) households had a designated toilet at home while the remaining 95 (95%) practiced open defecation at a site located on the slope of the hill behind the downhill colonies B and C (Figs. 2 and 3). Bore-wells were the main source of drinking water supply for 93 (93%) households. There were two deep borewells provided by the village administration and 17 shallow type bore-wells privately constructed by village residents. Five of these 17 (30%) shallow bore-wells were located in colonies B and C, on the downhill slope below the open defecation site. Plastic pipelines from the shallow wells were improperly installed with leakages at multiple points. There was no facility at source, for chlorination or any other mode of purification. Thirty households (30%) treated the water before consumption either by boiling or by use of candle filters. There was no routine drinking water surveillance in place by any authority for assessing the quality and fitness for drinking water. There was no residual chlorine found in any water samples. Three of four drinking water samples from bore-wells of most affected colonies (B and C) indicated faecal contamination by H2S field testing. There was heavy rainfall (average 65 mm in a day) from 22 to 24 June 2017. Prevention and control measures undertaken to contain the outbreak The village residents were discouraged from using shallow bore-well water and were provided with safe canned drinking water until all leakages were secured. Leakages in water supply from the bore-wells were identified and secured. Chlorine tablets were distributed for household level water disinfection. We informed the residents to avoid open defecation near drinking water sources and residential premises. Public health staff conducted health education daily to improve awareness among the villagers about water, sanitation, and hygiene. After active implementation of these control measures, cases declined rapidly in the village (Fig. 1). Discussion A rapid systematic epidemiological investigation of this outbreak identified water contamination points and likely mode of contamination. Based on these findings and our recommendations, the local health department instituted immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families. Effective implementation of public health measures limited the exposure of the community to contaminated water source resulting in rapid containment of the outbreak. Waterborne disease outbreaks tend to have cases spread over a time-period due to ongoing exposure to the contaminated water. In contrast, the pattern of epidemic curve in the present waterborne disease outbreak resembled that of food-borne with a point source exposure. Heavy rains contributed to the run-off of water from the open defecation site into the ground water of shallow wells located on slopes of hilly terrain resulting in heavy contamination and sudden rise of cases. Rapid control measures in the small village, implemented effectively within a short period of time, may have led to rapid decline of cases. The available epidemiological evidence also did not support generation of hypothesis of food-borne origin of the outbreak. In an outbreak reported among school children in Northern Greece in 2012, investigation revealed a waterborne viral gastroenteritis outbreak with a point source pattern, due to consumption of heavily contaminated water from a tap, which was not in use for two weeks during Christmas vacation [12]. Attack rate was high in this outbreak (37%), possibly due to exposure to high pathogen load subsequent to gross faecal contamination of water sources. In the absence of other alternative water sources, this tribal community was exclusively dependent on the contaminated water source for drinking, therefore exposing a large section of the community to risk. Geetha et al. analysed 32 diarrheal outbreaks in south India in non- tribal communities and reported lower attack rates varying from 0.6 to 21.5% [13]. However, tribal populations in India such as in Pedda-Gujjul-Thanda are marginalized with poor availability of WASH facilities [14]. This vulnerable tribal population continues to be at higher risk for ADD outbreaks with 27% not having access to safe drinking water and 75% of households not having toilets [15]. They need special assistance schemes from the government to enable them overcome poor accessibility to WASH facilities and secure healthy living [16]. Due to inadequate availability of communally managed safe public water points by the local authority, this community in Pedda-Gujjul-Thanda village was dependent on privately constructed shallow bore-wells for water supply. These are economical but likely to be unsafe. In this outbreak, open defecation site was present on the downhill slope in proximity to the residential premises and water resources, increasing the risk of drinking water contamination. Among the entire village population, 61% of ADD cases were attributable to drinking water from the ‘shallow downhill bore-wells’ (Population Attributable Fraction 61%), which was also evident from rapid outbreak containment following the elimination of exposure to this single risk factor. Since this exposure factor is amenable to long-term public health intervention, permanent elimination of shallow downhill borewells as water source was recommended, replacing them with properly secured deep bore-wells. Shallow bore-wells are known for their susceptibility to contamination from surface land-use activities [17, 18]. Studies have found levels of E. coli and enteric viruses to be high in shallow sources of ground water especially when they are in close proximity to polluting sources [19–21]. Consumption of ground water from shallow bore-wells with no purification facility increases the risk of diarrhea outbreaks manifold [22, 23]. A metaanalytic study of water-borne diarrheal disease outbreaks in China reported that 78 of 85 (92%) outbreaks (between year 1987 to 2014) were due to poor sanitary conditions of wells with lavatories/septic tanks nearby and lack of purification facilities [24]. In developed countries and urban areas of developing countries, as water supply and sanitation have improved dramatically over a period of time, such outbreaks were rarely reported in the recent past. The largest E. coli O157 outbreak in United States occurred in 1999 at a county fair (781 ill persons and 2 deaths) was due to groundwater source from a temporary unregulated well at the fairground [25]. Our findings have implication for India’s progress towards United Nation’s Sustainable Development Goal (SDG) 6 and India’s nation-wide campaign ‘Swachh Bharat Mission (SBM)’ to ensure availability and management of water and sanitation for all. SDG 6 aims at achieving universal access to basic sanitation service by 2030; and it has been reported that between 2000 and 2017, the proportion lacking even a basic sanitation service decreased from 44 to 27% [26]. SBM aims to achieve an open-defecation free status in rural areas through the construction of household-owned and community-owned toilets and establishing an accountable mechanism of monitoring toilet use. In 2015 in India, around 524 million (39%) practiced open defecation. However, under the SBM mission, due to increase in ‘households with toilets’ only 19 million (1.4%) practiced open defecation in January 2019 [5, 27]. There has also been a 71.58% increase in ‘households with toilets’ from October 2014 to October 2019 in rural areas of the Telangana state in India [27]. The tribal community initially obstructed the effective delivery of health care services; however, after involvement of the local stakeholders and tribal leaders, the acceptance towards medical treatment and community health services improved. Notwithstanding, most of the patients were still reluctant and did not consent for giving stool specimens for laboratory diagnosis. Establishing a rapport with the reticent tribal community was a major challenge faced by the outbreak investigation team. Lack of microbiological aetiology confirmation of the outbreak remained a limitation of the investigation due to limited stool samples and laboratory-capacity constraints of the remote area. Recognizing the pivotal importance of SDGs, national health policy of India (2017) has set the health-related cross-sectorial goal “access to safe water and sanitation to all by 2020†[28]. Greater political and financial commitment towards resource-limited remote tribal areas with effective community mobilization is required to accelerate the public health interventions to improve WASH and to prevent ADD outbreaks in the future. Conclusion This was a community-wide acute diarrheal disease outbreak with high village attack rate in a remote tribal village of Telangana with poor availability of safe water, sanitation and hygiene (WASH) facilities. A rapid and systematic epidemiological investigation identified drinking of faecal-contaminated water from the shallow borewells as the leading cause for this outbreak. These borewells were likely contaminated from runoff after rain from open defecation areas located on a downhill slope. Prompt and targeted public health action contained the number of cases.

What was the cause of the event?

95

A community-wide acute diarrheal disease outbreak associated with drinking contaminated water from shallow bore-wells in a tribal village

Abstract Background: In 2016, India reported 709 acute diarrheal disease (ADD) outbreaks (> 25% of all outbreaks). Tribal populations are at higher risk with 27% not having accessibility to safe drinking water and 75% households not having toilets. On June 26, 2017 Pedda-Gujjul-Thanda, a tribal village reported an acute diarrheal disease (ADD) outbreak. We investigated to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods: We defined a case as ≥3 loose stools within 24 h in Pedda-Gujjul-Thanda residents from June 24–30, 2017. We identified cases by reviewing hospital records and house-to-house survey. We conducted a retrospective cohort study and collected stool samples for culture. We assessed drinking water supply and sanitation practices and tested water samples for faecal-contamination. Results: We identified 191 cases (65% females) with median age 36 years (range 4–80 years) and no deaths. The attack-rate (AR) was 37% (191/512). Downhill colonies (located on slope of hilly terrains of the village) reported higher ARs (56%[136/243], p < 0.001) than others (20%[55/269]). Symptoms included diarrhea (100%), fever (17%), vomiting (16%) and abdominal pain (13%). Drinking water from five shallow bore-wells located in downhill colonies was significantly associated with illness (RR = 4.6, 95%CI = 3.4–6.1 and population attributable fraction 61%). In multivariate analysis, drinking water from the shallow bore-wells located in downhill colonies (aOR = 7.9, [95% CI =4.7– 13.2]), illiteracy (aOR =6, [95% CI = 3.6–10.1]), good hand-washing practice (aOR = 0.4, [95%CI = 0.2–0.7]) and household water treatment (aOR = 0.3, [95%CI = 0.2–0.5]) were significantly associated with illness. Two stool cultures were negative for Vibrio cholerae. Heavy rainfall was reported from June 22–24. Five of six water samples collected from shallow bore-wells located in downhill colonies were positive for faecal contamination. Conclusion: An ADD outbreak with high attack rate in a remote tribal village was associated with drinking water from shallow downhill bore-wells, likely contaminated via runoff from open defecation areas after heavy rains. Based on our recommendations, immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families, and as long-term public health measures construction of house-hold latrines and piped-water supply initiated. Keywords: Acute diarrheal disease, Outbreak, Bore-well, Tribal Introduction Globally there are an estimated 1.7 billion cases and 2.2 million deaths from acute diarrheal disease (ADD) every year [1]. In India, the burden is particularly high with more than 13.9 million cases reported in 2016 and 709 ADD outbreaks reported accounting to more than 25% of all outbreaks [2, 3]. Lack of access to safe drinking water and basic sanitation are the leading causes of ADD burden globally and in India. It is estimated that globally 58% of ADD deaths are attributed to inadequate drinking water, sanitation and hygiene [4]. The WHO/UNICEF Joint Monitoring Program for Water Supply, Sanitation and Hygiene (JMP) 2017 report revealed that 844 million people worldwide lack access to basic drinking-water service and 2.3 billion lack basic sanitation services, while 892 million still practiced open defecation [5]. The National Family Health Survey (NFHS-4, 2015–16) reported that in India only 52% of urban households and 18% of rural households have piped water supply, and the main source of water supply among rural households is bore-wells or tube-wells (51%). It has been estimated that 39% of households in India (54% among rural households) have no toilet facility and practicing open defecation [6]. The “indigenous†populations are socially, culturally and economically isolated and usually lack access to basic drinking-water and sanitation services. Therefore, they are vulnerable to ADD outbreaks and other emerging and re-emerging diseases [7]. The United Nations estimates that there are 370 million indigenous people existing across 90 countries of the world. They constitute 5% of the world population but 15% of the poorest [8]. India alone houses more than 705 such indigenous groups termed as Scheduled Tribes. As per the Census 2011, the total Scheduled Tribe population of India is 10.43 crore with a significant proportion of them living in rural areas [9]. On June 26, 2017, Kama-reddy district of Telangana state reported 55 ADD cases from the Pedda-GujjulThanda village. We conducted the outbreak investigation to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods Setting Pedda-Gujjul-Thanda village is a small tribal village with a total population of 563. The village is remotely located as an isolated community with a hilly terrain and is resource-limited with poor accessibility to sanitation and hygiene facilities. The nearest health care facility available for the residents is located at a distance of 10 km from the village. Case definition We defined a case as three or more loose stools within 24 h in a resident of the Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Case finding To find cases, we reviewed medical records of local health care facilities accessed by village residents in the nearby town. We conducted a medical camp in the village during the outbreak period for five days. We conducted a house-to-house survey in the village to find more cases, which are niether reported to health facility nor medical camp. Retrospective cohort study We conducted a retrospective cohort study to identify risk factors associated with illness. We defined the cohort as residents of Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Village resident was the unit of analysis. For data collection, we trained five teams of local paramedical staff. Using a pre-structured questionnaire, we collected data on demographic characteristics and risk factors related to drinking water, sanitation and hygiene. Good hand-washing practice was defined as reported washing of hands with soap and water every time after defecation and before eating. A bore-well less than 30-m-deep, as assessed from the records of village administration, was considered a shallow bore-well. Laboratory and environmental investigations Two stool samples were collected by the treating physician from admitted patients on the first day of hospital admission and transported to the state reference laboratory within two hours in Cary-Blair transport medium. The samples were cultured for Vibrio cholerae, Salmonella and Shigella on nutrient agar, MacConkey agar and deoxycholate citrate agar. Enteric pathogens were identified by biochemical reaction and by agglutination with anti-sera. We collected details of recent rainfall and conducted an environmental survey with household as sampling unit to assess drinking water, sanitation and hygiene practices. We assessed availability of residual chlorine in all village bore-wells and tested four of five bore-wells in the most affected colonies for faecal contamination by H2S method in field. Water was filled up to the ‘fill line’ of the sample bottle and incubated at room temperature (250–370 C) for 36–48 h and observed for colour change in the medium. A water sample was suspected to be contaminated with faecal matter, if it turned black [10, 11]. Because of limited supplies, we were unable to assess the fifth bore-well. Data analysis We analysed the data to describe the occurrence of cases over time, place, and person. We calculated relative risks (RR) with 95% confidence intervals (CI), population attributable risk percentages and conducted multiple logistic regression analysis with the dependent variables including consumption of shallow-downhill bore-well water, report of visible contaminants like mud in drinking water, illiteracy, household water treatment and good hand-washing practice. We used Epi Info version 7.2 for statistical analysis. Results Descriptive epidemiology We identified 191 ADD cases (65% females), with a village attack rate (AR) of 37% (191/512). The attack rate increased with age, with highest among > 60-year age group (55%) and lowest among children under-10 years (11%) (Table 1). No deaths were reported. In addition to diarrhea, cases presented with fever (17%), vomiting (16%) and abdominal pain (13%). 72% (138/191) cases reported to health care facilities and the medical camp conducted in the village. Among the 191 cases, 159 (83%) had mild illness treated with oral rehydration solution; 30 (16%) had moderate dehydration treated with intravenous fluids on out-patient basis, and 2 (1%) with severe dehydration were admitted in the district hospital for treatment with antibiotics (metronidazole and ciprofloxacin) and intravenous fluids. Cases started reported on June 26, 2017, with onset of symptoms from 24 June 2017. Maximum cases were reported on June 27, 2017, and no new cases were reported after June 30, 2017 (Fig. 1). The tribal population in the village had four sub-tribes namely Katroth, Badhawath, Nenawath, Baromath who resided in seven geographically demarcated colonies (labelled as A to G). Katroth sub-tribe resided in colonies A, B and G; Badhawath in colonies C and D; Nenawath in colony E and Baromath in colony F (Table 2). Colonies B and C had higher attack rates (65 and 47% respectively) as compared to other colonies (Fig. 2). Retrospective cohort study Among 563 village residents, 512 (91%) participated in the study. Among the 512 participants, median age was 28 years (range 1–80 years) with 52% females; 50% reported as illiterate with agriculture as the main source of livelihood for 76%. We analysed possible risk factors associated with ADD (Table 3). Drinking water from bore-well groundwater (vs canned water) was found significantly associated with ADD (RR = 12.7; 95% CI = 1.8–87.4). However, only 32 (6%) residents in the village used canned water and bore-well groundwater was the predominant source of water supply. Therefore, we analysed the water sources further, by location and type of bore-wells. Residents who used any of the five shallow bore-wells located downhill were significantly at higher risk (RR = 4.6; 95% CI = 3.4–6.1) and deep bore-wells were protective (RR = 0.4; 95% CI = 0.2–0.9). Report of visible contaminants like mud in drinking water (aOR = 4; 95% CI = 2.1–7.6) and illiteracy (aOR = 3.6; 95% CI = 3.5–10.1) were significantly associated with illness; and household water treatment (done either by boiling or use of candle filters) (aOR = 0.4; 95% CI = 0.2–0.7) and good hand-washing practice (aOR = 0.2; 95% CI = 0.1–0.5) were found protective. Laboratory and environmental results Stool samples collected from two hospitalized cases showed no growth for Vibrio cholerae, Salmonella and Shigella on culture. Among 110 households, 100 (91%) were available for environmental survey. Among the 100 houses surveyed, 79 (79%) were kutcha (low quality) type, made of mud, thatch and other low-quality material. Only 5 (5%) households had a designated toilet at home while the remaining 95 (95%) practiced open defecation at a site located on the slope of the hill behind the downhill colonies B and C (Figs. 2 and 3). Bore-wells were the main source of drinking water supply for 93 (93%) households. There were two deep borewells provided by the village administration and 17 shallow type bore-wells privately constructed by village residents. Five of these 17 (30%) shallow bore-wells were located in colonies B and C, on the downhill slope below the open defecation site. Plastic pipelines from the shallow wells were improperly installed with leakages at multiple points. There was no facility at source, for chlorination or any other mode of purification. Thirty households (30%) treated the water before consumption either by boiling or by use of candle filters. There was no routine drinking water surveillance in place by any authority for assessing the quality and fitness for drinking water. There was no residual chlorine found in any water samples. Three of four drinking water samples from bore-wells of most affected colonies (B and C) indicated faecal contamination by H2S field testing. There was heavy rainfall (average 65 mm in a day) from 22 to 24 June 2017. Prevention and control measures undertaken to contain the outbreak The village residents were discouraged from using shallow bore-well water and were provided with safe canned drinking water until all leakages were secured. Leakages in water supply from the bore-wells were identified and secured. Chlorine tablets were distributed for household level water disinfection. We informed the residents to avoid open defecation near drinking water sources and residential premises. Public health staff conducted health education daily to improve awareness among the villagers about water, sanitation, and hygiene. After active implementation of these control measures, cases declined rapidly in the village (Fig. 1). Discussion A rapid systematic epidemiological investigation of this outbreak identified water contamination points and likely mode of contamination. Based on these findings and our recommendations, the local health department instituted immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families. Effective implementation of public health measures limited the exposure of the community to contaminated water source resulting in rapid containment of the outbreak. Waterborne disease outbreaks tend to have cases spread over a time-period due to ongoing exposure to the contaminated water. In contrast, the pattern of epidemic curve in the present waterborne disease outbreak resembled that of food-borne with a point source exposure. Heavy rains contributed to the run-off of water from the open defecation site into the ground water of shallow wells located on slopes of hilly terrain resulting in heavy contamination and sudden rise of cases. Rapid control measures in the small village, implemented effectively within a short period of time, may have led to rapid decline of cases. The available epidemiological evidence also did not support generation of hypothesis of food-borne origin of the outbreak. In an outbreak reported among school children in Northern Greece in 2012, investigation revealed a waterborne viral gastroenteritis outbreak with a point source pattern, due to consumption of heavily contaminated water from a tap, which was not in use for two weeks during Christmas vacation [12]. Attack rate was high in this outbreak (37%), possibly due to exposure to high pathogen load subsequent to gross faecal contamination of water sources. In the absence of other alternative water sources, this tribal community was exclusively dependent on the contaminated water source for drinking, therefore exposing a large section of the community to risk. Geetha et al. analysed 32 diarrheal outbreaks in south India in non- tribal communities and reported lower attack rates varying from 0.6 to 21.5% [13]. However, tribal populations in India such as in Pedda-Gujjul-Thanda are marginalized with poor availability of WASH facilities [14]. This vulnerable tribal population continues to be at higher risk for ADD outbreaks with 27% not having access to safe drinking water and 75% of households not having toilets [15]. They need special assistance schemes from the government to enable them overcome poor accessibility to WASH facilities and secure healthy living [16]. Due to inadequate availability of communally managed safe public water points by the local authority, this community in Pedda-Gujjul-Thanda village was dependent on privately constructed shallow bore-wells for water supply. These are economical but likely to be unsafe. In this outbreak, open defecation site was present on the downhill slope in proximity to the residential premises and water resources, increasing the risk of drinking water contamination. Among the entire village population, 61% of ADD cases were attributable to drinking water from the ‘shallow downhill bore-wells’ (Population Attributable Fraction 61%), which was also evident from rapid outbreak containment following the elimination of exposure to this single risk factor. Since this exposure factor is amenable to long-term public health intervention, permanent elimination of shallow downhill borewells as water source was recommended, replacing them with properly secured deep bore-wells. Shallow bore-wells are known for their susceptibility to contamination from surface land-use activities [17, 18]. Studies have found levels of E. coli and enteric viruses to be high in shallow sources of ground water especially when they are in close proximity to polluting sources [19–21]. Consumption of ground water from shallow bore-wells with no purification facility increases the risk of diarrhea outbreaks manifold [22, 23]. A metaanalytic study of water-borne diarrheal disease outbreaks in China reported that 78 of 85 (92%) outbreaks (between year 1987 to 2014) were due to poor sanitary conditions of wells with lavatories/septic tanks nearby and lack of purification facilities [24]. In developed countries and urban areas of developing countries, as water supply and sanitation have improved dramatically over a period of time, such outbreaks were rarely reported in the recent past. The largest E. coli O157 outbreak in United States occurred in 1999 at a county fair (781 ill persons and 2 deaths) was due to groundwater source from a temporary unregulated well at the fairground [25]. Our findings have implication for India’s progress towards United Nation’s Sustainable Development Goal (SDG) 6 and India’s nation-wide campaign ‘Swachh Bharat Mission (SBM)’ to ensure availability and management of water and sanitation for all. SDG 6 aims at achieving universal access to basic sanitation service by 2030; and it has been reported that between 2000 and 2017, the proportion lacking even a basic sanitation service decreased from 44 to 27% [26]. SBM aims to achieve an open-defecation free status in rural areas through the construction of household-owned and community-owned toilets and establishing an accountable mechanism of monitoring toilet use. In 2015 in India, around 524 million (39%) practiced open defecation. However, under the SBM mission, due to increase in ‘households with toilets’ only 19 million (1.4%) practiced open defecation in January 2019 [5, 27]. There has also been a 71.58% increase in ‘households with toilets’ from October 2014 to October 2019 in rural areas of the Telangana state in India [27]. The tribal community initially obstructed the effective delivery of health care services; however, after involvement of the local stakeholders and tribal leaders, the acceptance towards medical treatment and community health services improved. Notwithstanding, most of the patients were still reluctant and did not consent for giving stool specimens for laboratory diagnosis. Establishing a rapport with the reticent tribal community was a major challenge faced by the outbreak investigation team. Lack of microbiological aetiology confirmation of the outbreak remained a limitation of the investigation due to limited stool samples and laboratory-capacity constraints of the remote area. Recognizing the pivotal importance of SDGs, national health policy of India (2017) has set the health-related cross-sectorial goal “access to safe water and sanitation to all by 2020†[28]. Greater political and financial commitment towards resource-limited remote tribal areas with effective community mobilization is required to accelerate the public health interventions to improve WASH and to prevent ADD outbreaks in the future. Conclusion This was a community-wide acute diarrheal disease outbreak with high village attack rate in a remote tribal village of Telangana with poor availability of safe water, sanitation and hygiene (WASH) facilities. A rapid and systematic epidemiological investigation identified drinking of faecal-contaminated water from the shallow borewells as the leading cause for this outbreak. These borewells were likely contaminated from runoff after rain from open defecation areas located on a downhill slope. Prompt and targeted public health action contained the number of cases.

What source started the event?

96

A community-wide acute diarrheal disease outbreak associated with drinking contaminated water from shallow bore-wells in a tribal village

Abstract Background: In 2016, India reported 709 acute diarrheal disease (ADD) outbreaks (> 25% of all outbreaks). Tribal populations are at higher risk with 27% not having accessibility to safe drinking water and 75% households not having toilets. On June 26, 2017 Pedda-Gujjul-Thanda, a tribal village reported an acute diarrheal disease (ADD) outbreak. We investigated to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods: We defined a case as ≥3 loose stools within 24 h in Pedda-Gujjul-Thanda residents from June 24–30, 2017. We identified cases by reviewing hospital records and house-to-house survey. We conducted a retrospective cohort study and collected stool samples for culture. We assessed drinking water supply and sanitation practices and tested water samples for faecal-contamination. Results: We identified 191 cases (65% females) with median age 36 years (range 4–80 years) and no deaths. The attack-rate (AR) was 37% (191/512). Downhill colonies (located on slope of hilly terrains of the village) reported higher ARs (56%[136/243], p < 0.001) than others (20%[55/269]). Symptoms included diarrhea (100%), fever (17%), vomiting (16%) and abdominal pain (13%). Drinking water from five shallow bore-wells located in downhill colonies was significantly associated with illness (RR = 4.6, 95%CI = 3.4–6.1 and population attributable fraction 61%). In multivariate analysis, drinking water from the shallow bore-wells located in downhill colonies (aOR = 7.9, [95% CI =4.7– 13.2]), illiteracy (aOR =6, [95% CI = 3.6–10.1]), good hand-washing practice (aOR = 0.4, [95%CI = 0.2–0.7]) and household water treatment (aOR = 0.3, [95%CI = 0.2–0.5]) were significantly associated with illness. Two stool cultures were negative for Vibrio cholerae. Heavy rainfall was reported from June 22–24. Five of six water samples collected from shallow bore-wells located in downhill colonies were positive for faecal contamination. Conclusion: An ADD outbreak with high attack rate in a remote tribal village was associated with drinking water from shallow downhill bore-wells, likely contaminated via runoff from open defecation areas after heavy rains. Based on our recommendations, immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families, and as long-term public health measures construction of house-hold latrines and piped-water supply initiated. Keywords: Acute diarrheal disease, Outbreak, Bore-well, Tribal Introduction Globally there are an estimated 1.7 billion cases and 2.2 million deaths from acute diarrheal disease (ADD) every year [1]. In India, the burden is particularly high with more than 13.9 million cases reported in 2016 and 709 ADD outbreaks reported accounting to more than 25% of all outbreaks [2, 3]. Lack of access to safe drinking water and basic sanitation are the leading causes of ADD burden globally and in India. It is estimated that globally 58% of ADD deaths are attributed to inadequate drinking water, sanitation and hygiene [4]. The WHO/UNICEF Joint Monitoring Program for Water Supply, Sanitation and Hygiene (JMP) 2017 report revealed that 844 million people worldwide lack access to basic drinking-water service and 2.3 billion lack basic sanitation services, while 892 million still practiced open defecation [5]. The National Family Health Survey (NFHS-4, 2015–16) reported that in India only 52% of urban households and 18% of rural households have piped water supply, and the main source of water supply among rural households is bore-wells or tube-wells (51%). It has been estimated that 39% of households in India (54% among rural households) have no toilet facility and practicing open defecation [6]. The “indigenous†populations are socially, culturally and economically isolated and usually lack access to basic drinking-water and sanitation services. Therefore, they are vulnerable to ADD outbreaks and other emerging and re-emerging diseases [7]. The United Nations estimates that there are 370 million indigenous people existing across 90 countries of the world. They constitute 5% of the world population but 15% of the poorest [8]. India alone houses more than 705 such indigenous groups termed as Scheduled Tribes. As per the Census 2011, the total Scheduled Tribe population of India is 10.43 crore with a significant proportion of them living in rural areas [9]. On June 26, 2017, Kama-reddy district of Telangana state reported 55 ADD cases from the Pedda-GujjulThanda village. We conducted the outbreak investigation to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods Setting Pedda-Gujjul-Thanda village is a small tribal village with a total population of 563. The village is remotely located as an isolated community with a hilly terrain and is resource-limited with poor accessibility to sanitation and hygiene facilities. The nearest health care facility available for the residents is located at a distance of 10 km from the village. Case definition We defined a case as three or more loose stools within 24 h in a resident of the Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Case finding To find cases, we reviewed medical records of local health care facilities accessed by village residents in the nearby town. We conducted a medical camp in the village during the outbreak period for five days. We conducted a house-to-house survey in the village to find more cases, which are niether reported to health facility nor medical camp. Retrospective cohort study We conducted a retrospective cohort study to identify risk factors associated with illness. We defined the cohort as residents of Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Village resident was the unit of analysis. For data collection, we trained five teams of local paramedical staff. Using a pre-structured questionnaire, we collected data on demographic characteristics and risk factors related to drinking water, sanitation and hygiene. Good hand-washing practice was defined as reported washing of hands with soap and water every time after defecation and before eating. A bore-well less than 30-m-deep, as assessed from the records of village administration, was considered a shallow bore-well. Laboratory and environmental investigations Two stool samples were collected by the treating physician from admitted patients on the first day of hospital admission and transported to the state reference laboratory within two hours in Cary-Blair transport medium. The samples were cultured for Vibrio cholerae, Salmonella and Shigella on nutrient agar, MacConkey agar and deoxycholate citrate agar. Enteric pathogens were identified by biochemical reaction and by agglutination with anti-sera. We collected details of recent rainfall and conducted an environmental survey with household as sampling unit to assess drinking water, sanitation and hygiene practices. We assessed availability of residual chlorine in all village bore-wells and tested four of five bore-wells in the most affected colonies for faecal contamination by H2S method in field. Water was filled up to the ‘fill line’ of the sample bottle and incubated at room temperature (250–370 C) for 36–48 h and observed for colour change in the medium. A water sample was suspected to be contaminated with faecal matter, if it turned black [10, 11]. Because of limited supplies, we were unable to assess the fifth bore-well. Data analysis We analysed the data to describe the occurrence of cases over time, place, and person. We calculated relative risks (RR) with 95% confidence intervals (CI), population attributable risk percentages and conducted multiple logistic regression analysis with the dependent variables including consumption of shallow-downhill bore-well water, report of visible contaminants like mud in drinking water, illiteracy, household water treatment and good hand-washing practice. We used Epi Info version 7.2 for statistical analysis. Results Descriptive epidemiology We identified 191 ADD cases (65% females), with a village attack rate (AR) of 37% (191/512). The attack rate increased with age, with highest among > 60-year age group (55%) and lowest among children under-10 years (11%) (Table 1). No deaths were reported. In addition to diarrhea, cases presented with fever (17%), vomiting (16%) and abdominal pain (13%). 72% (138/191) cases reported to health care facilities and the medical camp conducted in the village. Among the 191 cases, 159 (83%) had mild illness treated with oral rehydration solution; 30 (16%) had moderate dehydration treated with intravenous fluids on out-patient basis, and 2 (1%) with severe dehydration were admitted in the district hospital for treatment with antibiotics (metronidazole and ciprofloxacin) and intravenous fluids. Cases started reported on June 26, 2017, with onset of symptoms from 24 June 2017. Maximum cases were reported on June 27, 2017, and no new cases were reported after June 30, 2017 (Fig. 1). The tribal population in the village had four sub-tribes namely Katroth, Badhawath, Nenawath, Baromath who resided in seven geographically demarcated colonies (labelled as A to G). Katroth sub-tribe resided in colonies A, B and G; Badhawath in colonies C and D; Nenawath in colony E and Baromath in colony F (Table 2). Colonies B and C had higher attack rates (65 and 47% respectively) as compared to other colonies (Fig. 2). Retrospective cohort study Among 563 village residents, 512 (91%) participated in the study. Among the 512 participants, median age was 28 years (range 1–80 years) with 52% females; 50% reported as illiterate with agriculture as the main source of livelihood for 76%. We analysed possible risk factors associated with ADD (Table 3). Drinking water from bore-well groundwater (vs canned water) was found significantly associated with ADD (RR = 12.7; 95% CI = 1.8–87.4). However, only 32 (6%) residents in the village used canned water and bore-well groundwater was the predominant source of water supply. Therefore, we analysed the water sources further, by location and type of bore-wells. Residents who used any of the five shallow bore-wells located downhill were significantly at higher risk (RR = 4.6; 95% CI = 3.4–6.1) and deep bore-wells were protective (RR = 0.4; 95% CI = 0.2–0.9). Report of visible contaminants like mud in drinking water (aOR = 4; 95% CI = 2.1–7.6) and illiteracy (aOR = 3.6; 95% CI = 3.5–10.1) were significantly associated with illness; and household water treatment (done either by boiling or use of candle filters) (aOR = 0.4; 95% CI = 0.2–0.7) and good hand-washing practice (aOR = 0.2; 95% CI = 0.1–0.5) were found protective. Laboratory and environmental results Stool samples collected from two hospitalized cases showed no growth for Vibrio cholerae, Salmonella and Shigella on culture. Among 110 households, 100 (91%) were available for environmental survey. Among the 100 houses surveyed, 79 (79%) were kutcha (low quality) type, made of mud, thatch and other low-quality material. Only 5 (5%) households had a designated toilet at home while the remaining 95 (95%) practiced open defecation at a site located on the slope of the hill behind the downhill colonies B and C (Figs. 2 and 3). Bore-wells were the main source of drinking water supply for 93 (93%) households. There were two deep borewells provided by the village administration and 17 shallow type bore-wells privately constructed by village residents. Five of these 17 (30%) shallow bore-wells were located in colonies B and C, on the downhill slope below the open defecation site. Plastic pipelines from the shallow wells were improperly installed with leakages at multiple points. There was no facility at source, for chlorination or any other mode of purification. Thirty households (30%) treated the water before consumption either by boiling or by use of candle filters. There was no routine drinking water surveillance in place by any authority for assessing the quality and fitness for drinking water. There was no residual chlorine found in any water samples. Three of four drinking water samples from bore-wells of most affected colonies (B and C) indicated faecal contamination by H2S field testing. There was heavy rainfall (average 65 mm in a day) from 22 to 24 June 2017. Prevention and control measures undertaken to contain the outbreak The village residents were discouraged from using shallow bore-well water and were provided with safe canned drinking water until all leakages were secured. Leakages in water supply from the bore-wells were identified and secured. Chlorine tablets were distributed for household level water disinfection. We informed the residents to avoid open defecation near drinking water sources and residential premises. Public health staff conducted health education daily to improve awareness among the villagers about water, sanitation, and hygiene. After active implementation of these control measures, cases declined rapidly in the village (Fig. 1). Discussion A rapid systematic epidemiological investigation of this outbreak identified water contamination points and likely mode of contamination. Based on these findings and our recommendations, the local health department instituted immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families. Effective implementation of public health measures limited the exposure of the community to contaminated water source resulting in rapid containment of the outbreak. Waterborne disease outbreaks tend to have cases spread over a time-period due to ongoing exposure to the contaminated water. In contrast, the pattern of epidemic curve in the present waterborne disease outbreak resembled that of food-borne with a point source exposure. Heavy rains contributed to the run-off of water from the open defecation site into the ground water of shallow wells located on slopes of hilly terrain resulting in heavy contamination and sudden rise of cases. Rapid control measures in the small village, implemented effectively within a short period of time, may have led to rapid decline of cases. The available epidemiological evidence also did not support generation of hypothesis of food-borne origin of the outbreak. In an outbreak reported among school children in Northern Greece in 2012, investigation revealed a waterborne viral gastroenteritis outbreak with a point source pattern, due to consumption of heavily contaminated water from a tap, which was not in use for two weeks during Christmas vacation [12]. Attack rate was high in this outbreak (37%), possibly due to exposure to high pathogen load subsequent to gross faecal contamination of water sources. In the absence of other alternative water sources, this tribal community was exclusively dependent on the contaminated water source for drinking, therefore exposing a large section of the community to risk. Geetha et al. analysed 32 diarrheal outbreaks in south India in non- tribal communities and reported lower attack rates varying from 0.6 to 21.5% [13]. However, tribal populations in India such as in Pedda-Gujjul-Thanda are marginalized with poor availability of WASH facilities [14]. This vulnerable tribal population continues to be at higher risk for ADD outbreaks with 27% not having access to safe drinking water and 75% of households not having toilets [15]. They need special assistance schemes from the government to enable them overcome poor accessibility to WASH facilities and secure healthy living [16]. Due to inadequate availability of communally managed safe public water points by the local authority, this community in Pedda-Gujjul-Thanda village was dependent on privately constructed shallow bore-wells for water supply. These are economical but likely to be unsafe. In this outbreak, open defecation site was present on the downhill slope in proximity to the residential premises and water resources, increasing the risk of drinking water contamination. Among the entire village population, 61% of ADD cases were attributable to drinking water from the ‘shallow downhill bore-wells’ (Population Attributable Fraction 61%), which was also evident from rapid outbreak containment following the elimination of exposure to this single risk factor. Since this exposure factor is amenable to long-term public health intervention, permanent elimination of shallow downhill borewells as water source was recommended, replacing them with properly secured deep bore-wells. Shallow bore-wells are known for their susceptibility to contamination from surface land-use activities [17, 18]. Studies have found levels of E. coli and enteric viruses to be high in shallow sources of ground water especially when they are in close proximity to polluting sources [19–21]. Consumption of ground water from shallow bore-wells with no purification facility increases the risk of diarrhea outbreaks manifold [22, 23]. A metaanalytic study of water-borne diarrheal disease outbreaks in China reported that 78 of 85 (92%) outbreaks (between year 1987 to 2014) were due to poor sanitary conditions of wells with lavatories/septic tanks nearby and lack of purification facilities [24]. In developed countries and urban areas of developing countries, as water supply and sanitation have improved dramatically over a period of time, such outbreaks were rarely reported in the recent past. The largest E. coli O157 outbreak in United States occurred in 1999 at a county fair (781 ill persons and 2 deaths) was due to groundwater source from a temporary unregulated well at the fairground [25]. Our findings have implication for India’s progress towards United Nation’s Sustainable Development Goal (SDG) 6 and India’s nation-wide campaign ‘Swachh Bharat Mission (SBM)’ to ensure availability and management of water and sanitation for all. SDG 6 aims at achieving universal access to basic sanitation service by 2030; and it has been reported that between 2000 and 2017, the proportion lacking even a basic sanitation service decreased from 44 to 27% [26]. SBM aims to achieve an open-defecation free status in rural areas through the construction of household-owned and community-owned toilets and establishing an accountable mechanism of monitoring toilet use. In 2015 in India, around 524 million (39%) practiced open defecation. However, under the SBM mission, due to increase in ‘households with toilets’ only 19 million (1.4%) practiced open defecation in January 2019 [5, 27]. There has also been a 71.58% increase in ‘households with toilets’ from October 2014 to October 2019 in rural areas of the Telangana state in India [27]. The tribal community initially obstructed the effective delivery of health care services; however, after involvement of the local stakeholders and tribal leaders, the acceptance towards medical treatment and community health services improved. Notwithstanding, most of the patients were still reluctant and did not consent for giving stool specimens for laboratory diagnosis. Establishing a rapport with the reticent tribal community was a major challenge faced by the outbreak investigation team. Lack of microbiological aetiology confirmation of the outbreak remained a limitation of the investigation due to limited stool samples and laboratory-capacity constraints of the remote area. Recognizing the pivotal importance of SDGs, national health policy of India (2017) has set the health-related cross-sectorial goal “access to safe water and sanitation to all by 2020†[28]. Greater political and financial commitment towards resource-limited remote tribal areas with effective community mobilization is required to accelerate the public health interventions to improve WASH and to prevent ADD outbreaks in the future. Conclusion This was a community-wide acute diarrheal disease outbreak with high village attack rate in a remote tribal village of Telangana with poor availability of safe water, sanitation and hygiene (WASH) facilities. A rapid and systematic epidemiological investigation identified drinking of faecal-contaminated water from the shallow borewells as the leading cause for this outbreak. These borewells were likely contaminated from runoff after rain from open defecation areas located on a downhill slope. Prompt and targeted public health action contained the number of cases.

How was the event first detected?

97

A community-wide acute diarrheal disease outbreak associated with drinking contaminated water from shallow bore-wells in a tribal village

Abstract Background: In 2016, India reported 709 acute diarrheal disease (ADD) outbreaks (> 25% of all outbreaks). Tribal populations are at higher risk with 27% not having accessibility to safe drinking water and 75% households not having toilets. On June 26, 2017 Pedda-Gujjul-Thanda, a tribal village reported an acute diarrheal disease (ADD) outbreak. We investigated to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods: We defined a case as ≥3 loose stools within 24 h in Pedda-Gujjul-Thanda residents from June 24–30, 2017. We identified cases by reviewing hospital records and house-to-house survey. We conducted a retrospective cohort study and collected stool samples for culture. We assessed drinking water supply and sanitation practices and tested water samples for faecal-contamination. Results: We identified 191 cases (65% females) with median age 36 years (range 4–80 years) and no deaths. The attack-rate (AR) was 37% (191/512). Downhill colonies (located on slope of hilly terrains of the village) reported higher ARs (56%[136/243], p < 0.001) than others (20%[55/269]). Symptoms included diarrhea (100%), fever (17%), vomiting (16%) and abdominal pain (13%). Drinking water from five shallow bore-wells located in downhill colonies was significantly associated with illness (RR = 4.6, 95%CI = 3.4–6.1 and population attributable fraction 61%). In multivariate analysis, drinking water from the shallow bore-wells located in downhill colonies (aOR = 7.9, [95% CI =4.7– 13.2]), illiteracy (aOR =6, [95% CI = 3.6–10.1]), good hand-washing practice (aOR = 0.4, [95%CI = 0.2–0.7]) and household water treatment (aOR = 0.3, [95%CI = 0.2–0.5]) were significantly associated with illness. Two stool cultures were negative for Vibrio cholerae. Heavy rainfall was reported from June 22–24. Five of six water samples collected from shallow bore-wells located in downhill colonies were positive for faecal contamination. Conclusion: An ADD outbreak with high attack rate in a remote tribal village was associated with drinking water from shallow downhill bore-wells, likely contaminated via runoff from open defecation areas after heavy rains. Based on our recommendations, immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families, and as long-term public health measures construction of house-hold latrines and piped-water supply initiated. Keywords: Acute diarrheal disease, Outbreak, Bore-well, Tribal Introduction Globally there are an estimated 1.7 billion cases and 2.2 million deaths from acute diarrheal disease (ADD) every year [1]. In India, the burden is particularly high with more than 13.9 million cases reported in 2016 and 709 ADD outbreaks reported accounting to more than 25% of all outbreaks [2, 3]. Lack of access to safe drinking water and basic sanitation are the leading causes of ADD burden globally and in India. It is estimated that globally 58% of ADD deaths are attributed to inadequate drinking water, sanitation and hygiene [4]. The WHO/UNICEF Joint Monitoring Program for Water Supply, Sanitation and Hygiene (JMP) 2017 report revealed that 844 million people worldwide lack access to basic drinking-water service and 2.3 billion lack basic sanitation services, while 892 million still practiced open defecation [5]. The National Family Health Survey (NFHS-4, 2015–16) reported that in India only 52% of urban households and 18% of rural households have piped water supply, and the main source of water supply among rural households is bore-wells or tube-wells (51%). It has been estimated that 39% of households in India (54% among rural households) have no toilet facility and practicing open defecation [6]. The “indigenous†populations are socially, culturally and economically isolated and usually lack access to basic drinking-water and sanitation services. Therefore, they are vulnerable to ADD outbreaks and other emerging and re-emerging diseases [7]. The United Nations estimates that there are 370 million indigenous people existing across 90 countries of the world. They constitute 5% of the world population but 15% of the poorest [8]. India alone houses more than 705 such indigenous groups termed as Scheduled Tribes. As per the Census 2011, the total Scheduled Tribe population of India is 10.43 crore with a significant proportion of them living in rural areas [9]. On June 26, 2017, Kama-reddy district of Telangana state reported 55 ADD cases from the Pedda-GujjulThanda village. We conducted the outbreak investigation to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods Setting Pedda-Gujjul-Thanda village is a small tribal village with a total population of 563. The village is remotely located as an isolated community with a hilly terrain and is resource-limited with poor accessibility to sanitation and hygiene facilities. The nearest health care facility available for the residents is located at a distance of 10 km from the village. Case definition We defined a case as three or more loose stools within 24 h in a resident of the Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Case finding To find cases, we reviewed medical records of local health care facilities accessed by village residents in the nearby town. We conducted a medical camp in the village during the outbreak period for five days. We conducted a house-to-house survey in the village to find more cases, which are niether reported to health facility nor medical camp. Retrospective cohort study We conducted a retrospective cohort study to identify risk factors associated with illness. We defined the cohort as residents of Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Village resident was the unit of analysis. For data collection, we trained five teams of local paramedical staff. Using a pre-structured questionnaire, we collected data on demographic characteristics and risk factors related to drinking water, sanitation and hygiene. Good hand-washing practice was defined as reported washing of hands with soap and water every time after defecation and before eating. A bore-well less than 30-m-deep, as assessed from the records of village administration, was considered a shallow bore-well. Laboratory and environmental investigations Two stool samples were collected by the treating physician from admitted patients on the first day of hospital admission and transported to the state reference laboratory within two hours in Cary-Blair transport medium. The samples were cultured for Vibrio cholerae, Salmonella and Shigella on nutrient agar, MacConkey agar and deoxycholate citrate agar. Enteric pathogens were identified by biochemical reaction and by agglutination with anti-sera. We collected details of recent rainfall and conducted an environmental survey with household as sampling unit to assess drinking water, sanitation and hygiene practices. We assessed availability of residual chlorine in all village bore-wells and tested four of five bore-wells in the most affected colonies for faecal contamination by H2S method in field. Water was filled up to the ‘fill line’ of the sample bottle and incubated at room temperature (250–370 C) for 36–48 h and observed for colour change in the medium. A water sample was suspected to be contaminated with faecal matter, if it turned black [10, 11]. Because of limited supplies, we were unable to assess the fifth bore-well. Data analysis We analysed the data to describe the occurrence of cases over time, place, and person. We calculated relative risks (RR) with 95% confidence intervals (CI), population attributable risk percentages and conducted multiple logistic regression analysis with the dependent variables including consumption of shallow-downhill bore-well water, report of visible contaminants like mud in drinking water, illiteracy, household water treatment and good hand-washing practice. We used Epi Info version 7.2 for statistical analysis. Results Descriptive epidemiology We identified 191 ADD cases (65% females), with a village attack rate (AR) of 37% (191/512). The attack rate increased with age, with highest among > 60-year age group (55%) and lowest among children under-10 years (11%) (Table 1). No deaths were reported. In addition to diarrhea, cases presented with fever (17%), vomiting (16%) and abdominal pain (13%). 72% (138/191) cases reported to health care facilities and the medical camp conducted in the village. Among the 191 cases, 159 (83%) had mild illness treated with oral rehydration solution; 30 (16%) had moderate dehydration treated with intravenous fluids on out-patient basis, and 2 (1%) with severe dehydration were admitted in the district hospital for treatment with antibiotics (metronidazole and ciprofloxacin) and intravenous fluids. Cases started reported on June 26, 2017, with onset of symptoms from 24 June 2017. Maximum cases were reported on June 27, 2017, and no new cases were reported after June 30, 2017 (Fig. 1). The tribal population in the village had four sub-tribes namely Katroth, Badhawath, Nenawath, Baromath who resided in seven geographically demarcated colonies (labelled as A to G). Katroth sub-tribe resided in colonies A, B and G; Badhawath in colonies C and D; Nenawath in colony E and Baromath in colony F (Table 2). Colonies B and C had higher attack rates (65 and 47% respectively) as compared to other colonies (Fig. 2). Retrospective cohort study Among 563 village residents, 512 (91%) participated in the study. Among the 512 participants, median age was 28 years (range 1–80 years) with 52% females; 50% reported as illiterate with agriculture as the main source of livelihood for 76%. We analysed possible risk factors associated with ADD (Table 3). Drinking water from bore-well groundwater (vs canned water) was found significantly associated with ADD (RR = 12.7; 95% CI = 1.8–87.4). However, only 32 (6%) residents in the village used canned water and bore-well groundwater was the predominant source of water supply. Therefore, we analysed the water sources further, by location and type of bore-wells. Residents who used any of the five shallow bore-wells located downhill were significantly at higher risk (RR = 4.6; 95% CI = 3.4–6.1) and deep bore-wells were protective (RR = 0.4; 95% CI = 0.2–0.9). Report of visible contaminants like mud in drinking water (aOR = 4; 95% CI = 2.1–7.6) and illiteracy (aOR = 3.6; 95% CI = 3.5–10.1) were significantly associated with illness; and household water treatment (done either by boiling or use of candle filters) (aOR = 0.4; 95% CI = 0.2–0.7) and good hand-washing practice (aOR = 0.2; 95% CI = 0.1–0.5) were found protective. Laboratory and environmental results Stool samples collected from two hospitalized cases showed no growth for Vibrio cholerae, Salmonella and Shigella on culture. Among 110 households, 100 (91%) were available for environmental survey. Among the 100 houses surveyed, 79 (79%) were kutcha (low quality) type, made of mud, thatch and other low-quality material. Only 5 (5%) households had a designated toilet at home while the remaining 95 (95%) practiced open defecation at a site located on the slope of the hill behind the downhill colonies B and C (Figs. 2 and 3). Bore-wells were the main source of drinking water supply for 93 (93%) households. There were two deep borewells provided by the village administration and 17 shallow type bore-wells privately constructed by village residents. Five of these 17 (30%) shallow bore-wells were located in colonies B and C, on the downhill slope below the open defecation site. Plastic pipelines from the shallow wells were improperly installed with leakages at multiple points. There was no facility at source, for chlorination or any other mode of purification. Thirty households (30%) treated the water before consumption either by boiling or by use of candle filters. There was no routine drinking water surveillance in place by any authority for assessing the quality and fitness for drinking water. There was no residual chlorine found in any water samples. Three of four drinking water samples from bore-wells of most affected colonies (B and C) indicated faecal contamination by H2S field testing. There was heavy rainfall (average 65 mm in a day) from 22 to 24 June 2017. Prevention and control measures undertaken to contain the outbreak The village residents were discouraged from using shallow bore-well water and were provided with safe canned drinking water until all leakages were secured. Leakages in water supply from the bore-wells were identified and secured. Chlorine tablets were distributed for household level water disinfection. We informed the residents to avoid open defecation near drinking water sources and residential premises. Public health staff conducted health education daily to improve awareness among the villagers about water, sanitation, and hygiene. After active implementation of these control measures, cases declined rapidly in the village (Fig. 1). Discussion A rapid systematic epidemiological investigation of this outbreak identified water contamination points and likely mode of contamination. Based on these findings and our recommendations, the local health department instituted immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families. Effective implementation of public health measures limited the exposure of the community to contaminated water source resulting in rapid containment of the outbreak. Waterborne disease outbreaks tend to have cases spread over a time-period due to ongoing exposure to the contaminated water. In contrast, the pattern of epidemic curve in the present waterborne disease outbreak resembled that of food-borne with a point source exposure. Heavy rains contributed to the run-off of water from the open defecation site into the ground water of shallow wells located on slopes of hilly terrain resulting in heavy contamination and sudden rise of cases. Rapid control measures in the small village, implemented effectively within a short period of time, may have led to rapid decline of cases. The available epidemiological evidence also did not support generation of hypothesis of food-borne origin of the outbreak. In an outbreak reported among school children in Northern Greece in 2012, investigation revealed a waterborne viral gastroenteritis outbreak with a point source pattern, due to consumption of heavily contaminated water from a tap, which was not in use for two weeks during Christmas vacation [12]. Attack rate was high in this outbreak (37%), possibly due to exposure to high pathogen load subsequent to gross faecal contamination of water sources. In the absence of other alternative water sources, this tribal community was exclusively dependent on the contaminated water source for drinking, therefore exposing a large section of the community to risk. Geetha et al. analysed 32 diarrheal outbreaks in south India in non- tribal communities and reported lower attack rates varying from 0.6 to 21.5% [13]. However, tribal populations in India such as in Pedda-Gujjul-Thanda are marginalized with poor availability of WASH facilities [14]. This vulnerable tribal population continues to be at higher risk for ADD outbreaks with 27% not having access to safe drinking water and 75% of households not having toilets [15]. They need special assistance schemes from the government to enable them overcome poor accessibility to WASH facilities and secure healthy living [16]. Due to inadequate availability of communally managed safe public water points by the local authority, this community in Pedda-Gujjul-Thanda village was dependent on privately constructed shallow bore-wells for water supply. These are economical but likely to be unsafe. In this outbreak, open defecation site was present on the downhill slope in proximity to the residential premises and water resources, increasing the risk of drinking water contamination. Among the entire village population, 61% of ADD cases were attributable to drinking water from the ‘shallow downhill bore-wells’ (Population Attributable Fraction 61%), which was also evident from rapid outbreak containment following the elimination of exposure to this single risk factor. Since this exposure factor is amenable to long-term public health intervention, permanent elimination of shallow downhill borewells as water source was recommended, replacing them with properly secured deep bore-wells. Shallow bore-wells are known for their susceptibility to contamination from surface land-use activities [17, 18]. Studies have found levels of E. coli and enteric viruses to be high in shallow sources of ground water especially when they are in close proximity to polluting sources [19–21]. Consumption of ground water from shallow bore-wells with no purification facility increases the risk of diarrhea outbreaks manifold [22, 23]. A metaanalytic study of water-borne diarrheal disease outbreaks in China reported that 78 of 85 (92%) outbreaks (between year 1987 to 2014) were due to poor sanitary conditions of wells with lavatories/septic tanks nearby and lack of purification facilities [24]. In developed countries and urban areas of developing countries, as water supply and sanitation have improved dramatically over a period of time, such outbreaks were rarely reported in the recent past. The largest E. coli O157 outbreak in United States occurred in 1999 at a county fair (781 ill persons and 2 deaths) was due to groundwater source from a temporary unregulated well at the fairground [25]. Our findings have implication for India’s progress towards United Nation’s Sustainable Development Goal (SDG) 6 and India’s nation-wide campaign ‘Swachh Bharat Mission (SBM)’ to ensure availability and management of water and sanitation for all. SDG 6 aims at achieving universal access to basic sanitation service by 2030; and it has been reported that between 2000 and 2017, the proportion lacking even a basic sanitation service decreased from 44 to 27% [26]. SBM aims to achieve an open-defecation free status in rural areas through the construction of household-owned and community-owned toilets and establishing an accountable mechanism of monitoring toilet use. In 2015 in India, around 524 million (39%) practiced open defecation. However, under the SBM mission, due to increase in ‘households with toilets’ only 19 million (1.4%) practiced open defecation in January 2019 [5, 27]. There has also been a 71.58% increase in ‘households with toilets’ from October 2014 to October 2019 in rural areas of the Telangana state in India [27]. The tribal community initially obstructed the effective delivery of health care services; however, after involvement of the local stakeholders and tribal leaders, the acceptance towards medical treatment and community health services improved. Notwithstanding, most of the patients were still reluctant and did not consent for giving stool specimens for laboratory diagnosis. Establishing a rapport with the reticent tribal community was a major challenge faced by the outbreak investigation team. Lack of microbiological aetiology confirmation of the outbreak remained a limitation of the investigation due to limited stool samples and laboratory-capacity constraints of the remote area. Recognizing the pivotal importance of SDGs, national health policy of India (2017) has set the health-related cross-sectorial goal “access to safe water and sanitation to all by 2020†[28]. Greater political and financial commitment towards resource-limited remote tribal areas with effective community mobilization is required to accelerate the public health interventions to improve WASH and to prevent ADD outbreaks in the future. Conclusion This was a community-wide acute diarrheal disease outbreak with high village attack rate in a remote tribal village of Telangana with poor availability of safe water, sanitation and hygiene (WASH) facilities. A rapid and systematic epidemiological investigation identified drinking of faecal-contaminated water from the shallow borewells as the leading cause for this outbreak. These borewells were likely contaminated from runoff after rain from open defecation areas located on a downhill slope. Prompt and targeted public health action contained the number of cases.

How many people were ill?

98

A community-wide acute diarrheal disease outbreak associated with drinking contaminated water from shallow bore-wells in a tribal village

Abstract Background: In 2016, India reported 709 acute diarrheal disease (ADD) outbreaks (> 25% of all outbreaks). Tribal populations are at higher risk with 27% not having accessibility to safe drinking water and 75% households not having toilets. On June 26, 2017 Pedda-Gujjul-Thanda, a tribal village reported an acute diarrheal disease (ADD) outbreak. We investigated to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods: We defined a case as ≥3 loose stools within 24 h in Pedda-Gujjul-Thanda residents from June 24–30, 2017. We identified cases by reviewing hospital records and house-to-house survey. We conducted a retrospective cohort study and collected stool samples for culture. We assessed drinking water supply and sanitation practices and tested water samples for faecal-contamination. Results: We identified 191 cases (65% females) with median age 36 years (range 4–80 years) and no deaths. The attack-rate (AR) was 37% (191/512). Downhill colonies (located on slope of hilly terrains of the village) reported higher ARs (56%[136/243], p < 0.001) than others (20%[55/269]). Symptoms included diarrhea (100%), fever (17%), vomiting (16%) and abdominal pain (13%). Drinking water from five shallow bore-wells located in downhill colonies was significantly associated with illness (RR = 4.6, 95%CI = 3.4–6.1 and population attributable fraction 61%). In multivariate analysis, drinking water from the shallow bore-wells located in downhill colonies (aOR = 7.9, [95% CI =4.7– 13.2]), illiteracy (aOR =6, [95% CI = 3.6–10.1]), good hand-washing practice (aOR = 0.4, [95%CI = 0.2–0.7]) and household water treatment (aOR = 0.3, [95%CI = 0.2–0.5]) were significantly associated with illness. Two stool cultures were negative for Vibrio cholerae. Heavy rainfall was reported from June 22–24. Five of six water samples collected from shallow bore-wells located in downhill colonies were positive for faecal contamination. Conclusion: An ADD outbreak with high attack rate in a remote tribal village was associated with drinking water from shallow downhill bore-wells, likely contaminated via runoff from open defecation areas after heavy rains. Based on our recommendations, immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families, and as long-term public health measures construction of house-hold latrines and piped-water supply initiated. Keywords: Acute diarrheal disease, Outbreak, Bore-well, Tribal Introduction Globally there are an estimated 1.7 billion cases and 2.2 million deaths from acute diarrheal disease (ADD) every year [1]. In India, the burden is particularly high with more than 13.9 million cases reported in 2016 and 709 ADD outbreaks reported accounting to more than 25% of all outbreaks [2, 3]. Lack of access to safe drinking water and basic sanitation are the leading causes of ADD burden globally and in India. It is estimated that globally 58% of ADD deaths are attributed to inadequate drinking water, sanitation and hygiene [4]. The WHO/UNICEF Joint Monitoring Program for Water Supply, Sanitation and Hygiene (JMP) 2017 report revealed that 844 million people worldwide lack access to basic drinking-water service and 2.3 billion lack basic sanitation services, while 892 million still practiced open defecation [5]. The National Family Health Survey (NFHS-4, 2015–16) reported that in India only 52% of urban households and 18% of rural households have piped water supply, and the main source of water supply among rural households is bore-wells or tube-wells (51%). It has been estimated that 39% of households in India (54% among rural households) have no toilet facility and practicing open defecation [6]. The “indigenous†populations are socially, culturally and economically isolated and usually lack access to basic drinking-water and sanitation services. Therefore, they are vulnerable to ADD outbreaks and other emerging and re-emerging diseases [7]. The United Nations estimates that there are 370 million indigenous people existing across 90 countries of the world. They constitute 5% of the world population but 15% of the poorest [8]. India alone houses more than 705 such indigenous groups termed as Scheduled Tribes. As per the Census 2011, the total Scheduled Tribe population of India is 10.43 crore with a significant proportion of them living in rural areas [9]. On June 26, 2017, Kama-reddy district of Telangana state reported 55 ADD cases from the Pedda-GujjulThanda village. We conducted the outbreak investigation to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods Setting Pedda-Gujjul-Thanda village is a small tribal village with a total population of 563. The village is remotely located as an isolated community with a hilly terrain and is resource-limited with poor accessibility to sanitation and hygiene facilities. The nearest health care facility available for the residents is located at a distance of 10 km from the village. Case definition We defined a case as three or more loose stools within 24 h in a resident of the Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Case finding To find cases, we reviewed medical records of local health care facilities accessed by village residents in the nearby town. We conducted a medical camp in the village during the outbreak period for five days. We conducted a house-to-house survey in the village to find more cases, which are niether reported to health facility nor medical camp. Retrospective cohort study We conducted a retrospective cohort study to identify risk factors associated with illness. We defined the cohort as residents of Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Village resident was the unit of analysis. For data collection, we trained five teams of local paramedical staff. Using a pre-structured questionnaire, we collected data on demographic characteristics and risk factors related to drinking water, sanitation and hygiene. Good hand-washing practice was defined as reported washing of hands with soap and water every time after defecation and before eating. A bore-well less than 30-m-deep, as assessed from the records of village administration, was considered a shallow bore-well. Laboratory and environmental investigations Two stool samples were collected by the treating physician from admitted patients on the first day of hospital admission and transported to the state reference laboratory within two hours in Cary-Blair transport medium. The samples were cultured for Vibrio cholerae, Salmonella and Shigella on nutrient agar, MacConkey agar and deoxycholate citrate agar. Enteric pathogens were identified by biochemical reaction and by agglutination with anti-sera. We collected details of recent rainfall and conducted an environmental survey with household as sampling unit to assess drinking water, sanitation and hygiene practices. We assessed availability of residual chlorine in all village bore-wells and tested four of five bore-wells in the most affected colonies for faecal contamination by H2S method in field. Water was filled up to the ‘fill line’ of the sample bottle and incubated at room temperature (250–370 C) for 36–48 h and observed for colour change in the medium. A water sample was suspected to be contaminated with faecal matter, if it turned black [10, 11]. Because of limited supplies, we were unable to assess the fifth bore-well. Data analysis We analysed the data to describe the occurrence of cases over time, place, and person. We calculated relative risks (RR) with 95% confidence intervals (CI), population attributable risk percentages and conducted multiple logistic regression analysis with the dependent variables including consumption of shallow-downhill bore-well water, report of visible contaminants like mud in drinking water, illiteracy, household water treatment and good hand-washing practice. We used Epi Info version 7.2 for statistical analysis. Results Descriptive epidemiology We identified 191 ADD cases (65% females), with a village attack rate (AR) of 37% (191/512). The attack rate increased with age, with highest among > 60-year age group (55%) and lowest among children under-10 years (11%) (Table 1). No deaths were reported. In addition to diarrhea, cases presented with fever (17%), vomiting (16%) and abdominal pain (13%). 72% (138/191) cases reported to health care facilities and the medical camp conducted in the village. Among the 191 cases, 159 (83%) had mild illness treated with oral rehydration solution; 30 (16%) had moderate dehydration treated with intravenous fluids on out-patient basis, and 2 (1%) with severe dehydration were admitted in the district hospital for treatment with antibiotics (metronidazole and ciprofloxacin) and intravenous fluids. Cases started reported on June 26, 2017, with onset of symptoms from 24 June 2017. Maximum cases were reported on June 27, 2017, and no new cases were reported after June 30, 2017 (Fig. 1). The tribal population in the village had four sub-tribes namely Katroth, Badhawath, Nenawath, Baromath who resided in seven geographically demarcated colonies (labelled as A to G). Katroth sub-tribe resided in colonies A, B and G; Badhawath in colonies C and D; Nenawath in colony E and Baromath in colony F (Table 2). Colonies B and C had higher attack rates (65 and 47% respectively) as compared to other colonies (Fig. 2). Retrospective cohort study Among 563 village residents, 512 (91%) participated in the study. Among the 512 participants, median age was 28 years (range 1–80 years) with 52% females; 50% reported as illiterate with agriculture as the main source of livelihood for 76%. We analysed possible risk factors associated with ADD (Table 3). Drinking water from bore-well groundwater (vs canned water) was found significantly associated with ADD (RR = 12.7; 95% CI = 1.8–87.4). However, only 32 (6%) residents in the village used canned water and bore-well groundwater was the predominant source of water supply. Therefore, we analysed the water sources further, by location and type of bore-wells. Residents who used any of the five shallow bore-wells located downhill were significantly at higher risk (RR = 4.6; 95% CI = 3.4–6.1) and deep bore-wells were protective (RR = 0.4; 95% CI = 0.2–0.9). Report of visible contaminants like mud in drinking water (aOR = 4; 95% CI = 2.1–7.6) and illiteracy (aOR = 3.6; 95% CI = 3.5–10.1) were significantly associated with illness; and household water treatment (done either by boiling or use of candle filters) (aOR = 0.4; 95% CI = 0.2–0.7) and good hand-washing practice (aOR = 0.2; 95% CI = 0.1–0.5) were found protective. Laboratory and environmental results Stool samples collected from two hospitalized cases showed no growth for Vibrio cholerae, Salmonella and Shigella on culture. Among 110 households, 100 (91%) were available for environmental survey. Among the 100 houses surveyed, 79 (79%) were kutcha (low quality) type, made of mud, thatch and other low-quality material. Only 5 (5%) households had a designated toilet at home while the remaining 95 (95%) practiced open defecation at a site located on the slope of the hill behind the downhill colonies B and C (Figs. 2 and 3). Bore-wells were the main source of drinking water supply for 93 (93%) households. There were two deep borewells provided by the village administration and 17 shallow type bore-wells privately constructed by village residents. Five of these 17 (30%) shallow bore-wells were located in colonies B and C, on the downhill slope below the open defecation site. Plastic pipelines from the shallow wells were improperly installed with leakages at multiple points. There was no facility at source, for chlorination or any other mode of purification. Thirty households (30%) treated the water before consumption either by boiling or by use of candle filters. There was no routine drinking water surveillance in place by any authority for assessing the quality and fitness for drinking water. There was no residual chlorine found in any water samples. Three of four drinking water samples from bore-wells of most affected colonies (B and C) indicated faecal contamination by H2S field testing. There was heavy rainfall (average 65 mm in a day) from 22 to 24 June 2017. Prevention and control measures undertaken to contain the outbreak The village residents were discouraged from using shallow bore-well water and were provided with safe canned drinking water until all leakages were secured. Leakages in water supply from the bore-wells were identified and secured. Chlorine tablets were distributed for household level water disinfection. We informed the residents to avoid open defecation near drinking water sources and residential premises. Public health staff conducted health education daily to improve awareness among the villagers about water, sanitation, and hygiene. After active implementation of these control measures, cases declined rapidly in the village (Fig. 1). Discussion A rapid systematic epidemiological investigation of this outbreak identified water contamination points and likely mode of contamination. Based on these findings and our recommendations, the local health department instituted immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families. Effective implementation of public health measures limited the exposure of the community to contaminated water source resulting in rapid containment of the outbreak. Waterborne disease outbreaks tend to have cases spread over a time-period due to ongoing exposure to the contaminated water. In contrast, the pattern of epidemic curve in the present waterborne disease outbreak resembled that of food-borne with a point source exposure. Heavy rains contributed to the run-off of water from the open defecation site into the ground water of shallow wells located on slopes of hilly terrain resulting in heavy contamination and sudden rise of cases. Rapid control measures in the small village, implemented effectively within a short period of time, may have led to rapid decline of cases. The available epidemiological evidence also did not support generation of hypothesis of food-borne origin of the outbreak. In an outbreak reported among school children in Northern Greece in 2012, investigation revealed a waterborne viral gastroenteritis outbreak with a point source pattern, due to consumption of heavily contaminated water from a tap, which was not in use for two weeks during Christmas vacation [12]. Attack rate was high in this outbreak (37%), possibly due to exposure to high pathogen load subsequent to gross faecal contamination of water sources. In the absence of other alternative water sources, this tribal community was exclusively dependent on the contaminated water source for drinking, therefore exposing a large section of the community to risk. Geetha et al. analysed 32 diarrheal outbreaks in south India in non- tribal communities and reported lower attack rates varying from 0.6 to 21.5% [13]. However, tribal populations in India such as in Pedda-Gujjul-Thanda are marginalized with poor availability of WASH facilities [14]. This vulnerable tribal population continues to be at higher risk for ADD outbreaks with 27% not having access to safe drinking water and 75% of households not having toilets [15]. They need special assistance schemes from the government to enable them overcome poor accessibility to WASH facilities and secure healthy living [16]. Due to inadequate availability of communally managed safe public water points by the local authority, this community in Pedda-Gujjul-Thanda village was dependent on privately constructed shallow bore-wells for water supply. These are economical but likely to be unsafe. In this outbreak, open defecation site was present on the downhill slope in proximity to the residential premises and water resources, increasing the risk of drinking water contamination. Among the entire village population, 61% of ADD cases were attributable to drinking water from the ‘shallow downhill bore-wells’ (Population Attributable Fraction 61%), which was also evident from rapid outbreak containment following the elimination of exposure to this single risk factor. Since this exposure factor is amenable to long-term public health intervention, permanent elimination of shallow downhill borewells as water source was recommended, replacing them with properly secured deep bore-wells. Shallow bore-wells are known for their susceptibility to contamination from surface land-use activities [17, 18]. Studies have found levels of E. coli and enteric viruses to be high in shallow sources of ground water especially when they are in close proximity to polluting sources [19–21]. Consumption of ground water from shallow bore-wells with no purification facility increases the risk of diarrhea outbreaks manifold [22, 23]. A metaanalytic study of water-borne diarrheal disease outbreaks in China reported that 78 of 85 (92%) outbreaks (between year 1987 to 2014) were due to poor sanitary conditions of wells with lavatories/septic tanks nearby and lack of purification facilities [24]. In developed countries and urban areas of developing countries, as water supply and sanitation have improved dramatically over a period of time, such outbreaks were rarely reported in the recent past. The largest E. coli O157 outbreak in United States occurred in 1999 at a county fair (781 ill persons and 2 deaths) was due to groundwater source from a temporary unregulated well at the fairground [25]. Our findings have implication for India’s progress towards United Nation’s Sustainable Development Goal (SDG) 6 and India’s nation-wide campaign ‘Swachh Bharat Mission (SBM)’ to ensure availability and management of water and sanitation for all. SDG 6 aims at achieving universal access to basic sanitation service by 2030; and it has been reported that between 2000 and 2017, the proportion lacking even a basic sanitation service decreased from 44 to 27% [26]. SBM aims to achieve an open-defecation free status in rural areas through the construction of household-owned and community-owned toilets and establishing an accountable mechanism of monitoring toilet use. In 2015 in India, around 524 million (39%) practiced open defecation. However, under the SBM mission, due to increase in ‘households with toilets’ only 19 million (1.4%) practiced open defecation in January 2019 [5, 27]. There has also been a 71.58% increase in ‘households with toilets’ from October 2014 to October 2019 in rural areas of the Telangana state in India [27]. The tribal community initially obstructed the effective delivery of health care services; however, after involvement of the local stakeholders and tribal leaders, the acceptance towards medical treatment and community health services improved. Notwithstanding, most of the patients were still reluctant and did not consent for giving stool specimens for laboratory diagnosis. Establishing a rapport with the reticent tribal community was a major challenge faced by the outbreak investigation team. Lack of microbiological aetiology confirmation of the outbreak remained a limitation of the investigation due to limited stool samples and laboratory-capacity constraints of the remote area. Recognizing the pivotal importance of SDGs, national health policy of India (2017) has set the health-related cross-sectorial goal “access to safe water and sanitation to all by 2020†[28]. Greater political and financial commitment towards resource-limited remote tribal areas with effective community mobilization is required to accelerate the public health interventions to improve WASH and to prevent ADD outbreaks in the future. Conclusion This was a community-wide acute diarrheal disease outbreak with high village attack rate in a remote tribal village of Telangana with poor availability of safe water, sanitation and hygiene (WASH) facilities. A rapid and systematic epidemiological investigation identified drinking of faecal-contaminated water from the shallow borewells as the leading cause for this outbreak. These borewells were likely contaminated from runoff after rain from open defecation areas located on a downhill slope. Prompt and targeted public health action contained the number of cases.

How many people were hospitalized?

99

A community-wide acute diarrheal disease outbreak associated with drinking contaminated water from shallow bore-wells in a tribal village

Abstract Background: In 2016, India reported 709 acute diarrheal disease (ADD) outbreaks (> 25% of all outbreaks). Tribal populations are at higher risk with 27% not having accessibility to safe drinking water and 75% households not having toilets. On June 26, 2017 Pedda-Gujjul-Thanda, a tribal village reported an acute diarrheal disease (ADD) outbreak. We investigated to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods: We defined a case as ≥3 loose stools within 24 h in Pedda-Gujjul-Thanda residents from June 24–30, 2017. We identified cases by reviewing hospital records and house-to-house survey. We conducted a retrospective cohort study and collected stool samples for culture. We assessed drinking water supply and sanitation practices and tested water samples for faecal-contamination. Results: We identified 191 cases (65% females) with median age 36 years (range 4–80 years) and no deaths. The attack-rate (AR) was 37% (191/512). Downhill colonies (located on slope of hilly terrains of the village) reported higher ARs (56%[136/243], p < 0.001) than others (20%[55/269]). Symptoms included diarrhea (100%), fever (17%), vomiting (16%) and abdominal pain (13%). Drinking water from five shallow bore-wells located in downhill colonies was significantly associated with illness (RR = 4.6, 95%CI = 3.4–6.1 and population attributable fraction 61%). In multivariate analysis, drinking water from the shallow bore-wells located in downhill colonies (aOR = 7.9, [95% CI =4.7– 13.2]), illiteracy (aOR =6, [95% CI = 3.6–10.1]), good hand-washing practice (aOR = 0.4, [95%CI = 0.2–0.7]) and household water treatment (aOR = 0.3, [95%CI = 0.2–0.5]) were significantly associated with illness. Two stool cultures were negative for Vibrio cholerae. Heavy rainfall was reported from June 22–24. Five of six water samples collected from shallow bore-wells located in downhill colonies were positive for faecal contamination. Conclusion: An ADD outbreak with high attack rate in a remote tribal village was associated with drinking water from shallow downhill bore-wells, likely contaminated via runoff from open defecation areas after heavy rains. Based on our recommendations, immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families, and as long-term public health measures construction of house-hold latrines and piped-water supply initiated. Keywords: Acute diarrheal disease, Outbreak, Bore-well, Tribal Introduction Globally there are an estimated 1.7 billion cases and 2.2 million deaths from acute diarrheal disease (ADD) every year [1]. In India, the burden is particularly high with more than 13.9 million cases reported in 2016 and 709 ADD outbreaks reported accounting to more than 25% of all outbreaks [2, 3]. Lack of access to safe drinking water and basic sanitation are the leading causes of ADD burden globally and in India. It is estimated that globally 58% of ADD deaths are attributed to inadequate drinking water, sanitation and hygiene [4]. The WHO/UNICEF Joint Monitoring Program for Water Supply, Sanitation and Hygiene (JMP) 2017 report revealed that 844 million people worldwide lack access to basic drinking-water service and 2.3 billion lack basic sanitation services, while 892 million still practiced open defecation [5]. The National Family Health Survey (NFHS-4, 2015–16) reported that in India only 52% of urban households and 18% of rural households have piped water supply, and the main source of water supply among rural households is bore-wells or tube-wells (51%). It has been estimated that 39% of households in India (54% among rural households) have no toilet facility and practicing open defecation [6]. The “indigenous†populations are socially, culturally and economically isolated and usually lack access to basic drinking-water and sanitation services. Therefore, they are vulnerable to ADD outbreaks and other emerging and re-emerging diseases [7]. The United Nations estimates that there are 370 million indigenous people existing across 90 countries of the world. They constitute 5% of the world population but 15% of the poorest [8]. India alone houses more than 705 such indigenous groups termed as Scheduled Tribes. As per the Census 2011, the total Scheduled Tribe population of India is 10.43 crore with a significant proportion of them living in rural areas [9]. On June 26, 2017, Kama-reddy district of Telangana state reported 55 ADD cases from the Pedda-GujjulThanda village. We conducted the outbreak investigation to describe the epidemiology, identify risk factors, and provide evidence-based recommendations. Methods Setting Pedda-Gujjul-Thanda village is a small tribal village with a total population of 563. The village is remotely located as an isolated community with a hilly terrain and is resource-limited with poor accessibility to sanitation and hygiene facilities. The nearest health care facility available for the residents is located at a distance of 10 km from the village. Case definition We defined a case as three or more loose stools within 24 h in a resident of the Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Case finding To find cases, we reviewed medical records of local health care facilities accessed by village residents in the nearby town. We conducted a medical camp in the village during the outbreak period for five days. We conducted a house-to-house survey in the village to find more cases, which are niether reported to health facility nor medical camp. Retrospective cohort study We conducted a retrospective cohort study to identify risk factors associated with illness. We defined the cohort as residents of Pedda-Gujjul-Thanda village from June 22, 2017 to July 2, 2018. Village resident was the unit of analysis. For data collection, we trained five teams of local paramedical staff. Using a pre-structured questionnaire, we collected data on demographic characteristics and risk factors related to drinking water, sanitation and hygiene. Good hand-washing practice was defined as reported washing of hands with soap and water every time after defecation and before eating. A bore-well less than 30-m-deep, as assessed from the records of village administration, was considered a shallow bore-well. Laboratory and environmental investigations Two stool samples were collected by the treating physician from admitted patients on the first day of hospital admission and transported to the state reference laboratory within two hours in Cary-Blair transport medium. The samples were cultured for Vibrio cholerae, Salmonella and Shigella on nutrient agar, MacConkey agar and deoxycholate citrate agar. Enteric pathogens were identified by biochemical reaction and by agglutination with anti-sera. We collected details of recent rainfall and conducted an environmental survey with household as sampling unit to assess drinking water, sanitation and hygiene practices. We assessed availability of residual chlorine in all village bore-wells and tested four of five bore-wells in the most affected colonies for faecal contamination by H2S method in field. Water was filled up to the ‘fill line’ of the sample bottle and incubated at room temperature (250–370 C) for 36–48 h and observed for colour change in the medium. A water sample was suspected to be contaminated with faecal matter, if it turned black [10, 11]. Because of limited supplies, we were unable to assess the fifth bore-well. Data analysis We analysed the data to describe the occurrence of cases over time, place, and person. We calculated relative risks (RR) with 95% confidence intervals (CI), population attributable risk percentages and conducted multiple logistic regression analysis with the dependent variables including consumption of shallow-downhill bore-well water, report of visible contaminants like mud in drinking water, illiteracy, household water treatment and good hand-washing practice. We used Epi Info version 7.2 for statistical analysis. Results Descriptive epidemiology We identified 191 ADD cases (65% females), with a village attack rate (AR) of 37% (191/512). The attack rate increased with age, with highest among > 60-year age group (55%) and lowest among children under-10 years (11%) (Table 1). No deaths were reported. In addition to diarrhea, cases presented with fever (17%), vomiting (16%) and abdominal pain (13%). 72% (138/191) cases reported to health care facilities and the medical camp conducted in the village. Among the 191 cases, 159 (83%) had mild illness treated with oral rehydration solution; 30 (16%) had moderate dehydration treated with intravenous fluids on out-patient basis, and 2 (1%) with severe dehydration were admitted in the district hospital for treatment with antibiotics (metronidazole and ciprofloxacin) and intravenous fluids. Cases started reported on June 26, 2017, with onset of symptoms from 24 June 2017. Maximum cases were reported on June 27, 2017, and no new cases were reported after June 30, 2017 (Fig. 1). The tribal population in the village had four sub-tribes namely Katroth, Badhawath, Nenawath, Baromath who resided in seven geographically demarcated colonies (labelled as A to G). Katroth sub-tribe resided in colonies A, B and G; Badhawath in colonies C and D; Nenawath in colony E and Baromath in colony F (Table 2). Colonies B and C had higher attack rates (65 and 47% respectively) as compared to other colonies (Fig. 2). Retrospective cohort study Among 563 village residents, 512 (91%) participated in the study. Among the 512 participants, median age was 28 years (range 1–80 years) with 52% females; 50% reported as illiterate with agriculture as the main source of livelihood for 76%. We analysed possible risk factors associated with ADD (Table 3). Drinking water from bore-well groundwater (vs canned water) was found significantly associated with ADD (RR = 12.7; 95% CI = 1.8–87.4). However, only 32 (6%) residents in the village used canned water and bore-well groundwater was the predominant source of water supply. Therefore, we analysed the water sources further, by location and type of bore-wells. Residents who used any of the five shallow bore-wells located downhill were significantly at higher risk (RR = 4.6; 95% CI = 3.4–6.1) and deep bore-wells were protective (RR = 0.4; 95% CI = 0.2–0.9). Report of visible contaminants like mud in drinking water (aOR = 4; 95% CI = 2.1–7.6) and illiteracy (aOR = 3.6; 95% CI = 3.5–10.1) were significantly associated with illness; and household water treatment (done either by boiling or use of candle filters) (aOR = 0.4; 95% CI = 0.2–0.7) and good hand-washing practice (aOR = 0.2; 95% CI = 0.1–0.5) were found protective. Laboratory and environmental results Stool samples collected from two hospitalized cases showed no growth for Vibrio cholerae, Salmonella and Shigella on culture. Among 110 households, 100 (91%) were available for environmental survey. Among the 100 houses surveyed, 79 (79%) were kutcha (low quality) type, made of mud, thatch and other low-quality material. Only 5 (5%) households had a designated toilet at home while the remaining 95 (95%) practiced open defecation at a site located on the slope of the hill behind the downhill colonies B and C (Figs. 2 and 3). Bore-wells were the main source of drinking water supply for 93 (93%) households. There were two deep borewells provided by the village administration and 17 shallow type bore-wells privately constructed by village residents. Five of these 17 (30%) shallow bore-wells were located in colonies B and C, on the downhill slope below the open defecation site. Plastic pipelines from the shallow wells were improperly installed with leakages at multiple points. There was no facility at source, for chlorination or any other mode of purification. Thirty households (30%) treated the water before consumption either by boiling or by use of candle filters. There was no routine drinking water surveillance in place by any authority for assessing the quality and fitness for drinking water. There was no residual chlorine found in any water samples. Three of four drinking water samples from bore-wells of most affected colonies (B and C) indicated faecal contamination by H2S field testing. There was heavy rainfall (average 65 mm in a day) from 22 to 24 June 2017. Prevention and control measures undertaken to contain the outbreak The village residents were discouraged from using shallow bore-well water and were provided with safe canned drinking water until all leakages were secured. Leakages in water supply from the bore-wells were identified and secured. Chlorine tablets were distributed for household level water disinfection. We informed the residents to avoid open defecation near drinking water sources and residential premises. Public health staff conducted health education daily to improve awareness among the villagers about water, sanitation, and hygiene. After active implementation of these control measures, cases declined rapidly in the village (Fig. 1). Discussion A rapid systematic epidemiological investigation of this outbreak identified water contamination points and likely mode of contamination. Based on these findings and our recommendations, the local health department instituted immediate public health actions including repair of leakages at contaminated water sources and alternative supply of purified canned drinking water to families. Effective implementation of public health measures limited the exposure of the community to contaminated water source resulting in rapid containment of the outbreak. Waterborne disease outbreaks tend to have cases spread over a time-period due to ongoing exposure to the contaminated water. In contrast, the pattern of epidemic curve in the present waterborne disease outbreak resembled that of food-borne with a point source exposure. Heavy rains contributed to the run-off of water from the open defecation site into the ground water of shallow wells located on slopes of hilly terrain resulting in heavy contamination and sudden rise of cases. Rapid control measures in the small village, implemented effectively within a short period of time, may have led to rapid decline of cases. The available epidemiological evidence also did not support generation of hypothesis of food-borne origin of the outbreak. In an outbreak reported among school children in Northern Greece in 2012, investigation revealed a waterborne viral gastroenteritis outbreak with a point source pattern, due to consumption of heavily contaminated water from a tap, which was not in use for two weeks during Christmas vacation [12]. Attack rate was high in this outbreak (37%), possibly due to exposure to high pathogen load subsequent to gross faecal contamination of water sources. In the absence of other alternative water sources, this tribal community was exclusively dependent on the contaminated water source for drinking, therefore exposing a large section of the community to risk. Geetha et al. analysed 32 diarrheal outbreaks in south India in non- tribal communities and reported lower attack rates varying from 0.6 to 21.5% [13]. However, tribal populations in India such as in Pedda-Gujjul-Thanda are marginalized with poor availability of WASH facilities [14]. This vulnerable tribal population continues to be at higher risk for ADD outbreaks with 27% not having access to safe drinking water and 75% of households not having toilets [15]. They need special assistance schemes from the government to enable them overcome poor accessibility to WASH facilities and secure healthy living [16]. Due to inadequate availability of communally managed safe public water points by the local authority, this community in Pedda-Gujjul-Thanda village was dependent on privately constructed shallow bore-wells for water supply. These are economical but likely to be unsafe. In this outbreak, open defecation site was present on the downhill slope in proximity to the residential premises and water resources, increasing the risk of drinking water contamination. Among the entire village population, 61% of ADD cases were attributable to drinking water from the ‘shallow downhill bore-wells’ (Population Attributable Fraction 61%), which was also evident from rapid outbreak containment following the elimination of exposure to this single risk factor. Since this exposure factor is amenable to long-term public health intervention, permanent elimination of shallow downhill borewells as water source was recommended, replacing them with properly secured deep bore-wells. Shallow bore-wells are known for their susceptibility to contamination from surface land-use activities [17, 18]. Studies have found levels of E. coli and enteric viruses to be high in shallow sources of ground water especially when they are in close proximity to polluting sources [19–21]. Consumption of ground water from shallow bore-wells with no purification facility increases the risk of diarrhea outbreaks manifold [22, 23]. A metaanalytic study of water-borne diarrheal disease outbreaks in China reported that 78 of 85 (92%) outbreaks (between year 1987 to 2014) were due to poor sanitary conditions of wells with lavatories/septic tanks nearby and lack of purification facilities [24]. In developed countries and urban areas of developing countries, as water supply and sanitation have improved dramatically over a period of time, such outbreaks were rarely reported in the recent past. The largest E. coli O157 outbreak in United States occurred in 1999 at a county fair (781 ill persons and 2 deaths) was due to groundwater source from a temporary unregulated well at the fairground [25]. Our findings have implication for India’s progress towards United Nation’s Sustainable Development Goal (SDG) 6 and India’s nation-wide campaign ‘Swachh Bharat Mission (SBM)’ to ensure availability and management of water and sanitation for all. SDG 6 aims at achieving universal access to basic sanitation service by 2030; and it has been reported that between 2000 and 2017, the proportion lacking even a basic sanitation service decreased from 44 to 27% [26]. SBM aims to achieve an open-defecation free status in rural areas through the construction of household-owned and community-owned toilets and establishing an accountable mechanism of monitoring toilet use. In 2015 in India, around 524 million (39%) practiced open defecation. However, under the SBM mission, due to increase in ‘households with toilets’ only 19 million (1.4%) practiced open defecation in January 2019 [5, 27]. There has also been a 71.58% increase in ‘households with toilets’ from October 2014 to October 2019 in rural areas of the Telangana state in India [27]. The tribal community initially obstructed the effective delivery of health care services; however, after involvement of the local stakeholders and tribal leaders, the acceptance towards medical treatment and community health services improved. Notwithstanding, most of the patients were still reluctant and did not consent for giving stool specimens for laboratory diagnosis. Establishing a rapport with the reticent tribal community was a major challenge faced by the outbreak investigation team. Lack of microbiological aetiology confirmation of the outbreak remained a limitation of the investigation due to limited stool samples and laboratory-capacity constraints of the remote area. Recognizing the pivotal importance of SDGs, national health policy of India (2017) has set the health-related cross-sectorial goal “access to safe water and sanitation to all by 2020†[28]. Greater political and financial commitment towards resource-limited remote tribal areas with effective community mobilization is required to accelerate the public health interventions to improve WASH and to prevent ADD outbreaks in the future. Conclusion This was a community-wide acute diarrheal disease outbreak with high village attack rate in a remote tribal village of Telangana with poor availability of safe water, sanitation and hygiene (WASH) facilities. A rapid and systematic epidemiological investigation identified drinking of faecal-contaminated water from the shallow borewells as the leading cause for this outbreak. These borewells were likely contaminated from runoff after rain from open defecation areas located on a downhill slope. Prompt and targeted public health action contained the number of cases.

How many people were dead?