metadata

license: other
license_name: common-crawl
license_link: LICENSE
task_categories:
  - text-generation
language:
  - en
pretty_name: Clinical Guidelines
size_categories:
  - 10K<n<100K

Clinical Guidelines

The Clinical Guidelines corpus is a new dataset of 46,649 clinical practice guidelines from 16 high-quality online medical sources. This dataset serves as a crucial component of the original training corpus of the Meditron Large Language Model (LLM). We publicly release a subset of 35,733 articles from our Guidelines corpus, extracted from 8 of 16 sources that allow content redistribution, namely CCO, CDC, CMA, ICRC, NICE, SPOR, WHO and WikiDoc.

You can scrape and clean all 16 guideline sources using our code in epfLLM/meditron.

Source	Full Name	Tag	Guidelines	Words	Audience	Country	Released
AAFP	American Academy of Family Physicians	`aafp`	50	9.4K	Doctor	USA	No
CCO	Cancer Care Ontario	`cco`	87	199K	Doctor	Canada	Yes
CDC	Center for Disease Control and Prevention	`cdc`	621	6.7M	Doctor	USA	Yes
CMA	Canadian Medical Association	`cma`	431	1.7M	Doctor	Canada	Yes
CPS	Canadian Paediatric Society	`cps`	54	133K	Doctor	Canada	No
drugs.com	Drugs.com	`drugs`	6548	4.1M	Both	NZ	No
GuidelineCentral	GuidelineCentral	`gc`	1029	1M	Doctor	Mix	No
ICRC	International Committee of the Red Cross	`icrc`	49	1.2M	Doctor	Switzerland	Yes
IDSA	Infectious Diseases Society of America	`idsa`	47	646K	Doctor	USA	No
MAGIC	Making GRADE The Irresistible Choice	`magic`	52	415K	Doctor	Mix	No
MayoClinic	MayoClinic	`mayo`	1100	2.2M	Patient	USA	No
NICE	National Institute for Health and Care Excellence	`nice`	1656	8.1M	Doctor	UK	Yes
RCH	Royal Children's Hospital Melbourne	`rch`	384	410K	Doctor	Australia	No
SPOR	Strategy for Patient-Oriented Research	`spor`	217	1.1M	Doctor	Canada	Yes
WHO	World Health Organization	`who`	223	3.1M	Both	Switzerland	Yes
WikiDoc	WikiDoc	`wikidoc`	33058	34M	Both	International	Yes

Dataset Details

Dataset Description

Curated by: EPFL LLM Team
Funded by: [More Information Needed]
Language(s): English only
License: Common Crawl Foundation Terms of Use
Knowledge Cutoff: August 2023

Dataset Sources

Repository: epfLLM/meditron
Paper: MediTron-70B: Scaling Medical Pretraining for Large Language Models

Uses

Direct Use

The dataset is intended for use in tasks related to text generation, specifically in the context of clinical practice guidelines. It can be employed for training language models and other natural language processing applications within the healthcare domain.

Out-of-Scope Use

[More Information Needed]

Dataset Structure

Each row of the dataset represents one clinical practice guideline article, and consists of the following dataset fields (all strings):

Field	Description	Sources with field
`id`	Unique identifier for each article	All
`source`	Source tag (`cco`, `cdc`, `cma`, `icrc`, `nice`, `spor`, `who` or `wikidoc`)	All
`title`	Title of the article	CMA, NICE & WikiDoc only
`url`	URL of the article	NICE, WikiDoc only
`raw_text`	Unprocessed scraped article text	All
`clean_text`	Cleaned and formatted article text	All
`overview`	Short summary of the article	NICE only

Dataset Creation

Curation Rationale

The dataset was curated to provide a high-quality collection of clinical practice guidelines (CPGs) for the medical training of LLMs. Our Clinical Guidelines corpus comprises 46,469 articles from 16 globally recognized sources for clinician and patient-directed guidance across high and low-resource settings, multiple medical domains (internal medicine, pediatrics, oncology, infectious disease, etc.) and multiple geographical locations.

Clinical practice guidelines are rigorously researched frameworks designed to guide healthcare practitioners and patients in making evidence-based decisions regarding diagnosis, treatment, and management. They are compiled through a systematic process of collaborative consensus between experts to establish recommendations from the latest evidence on best practices that would maximize benefit in light of practical concerns such as available resources and context. As a super-synthesis of meta-analyses, they sit atop the evidence pyramid and form the basis of actionable evidence-based practice. CPGs are produced at various geographic and organizational granularities, ranging from global to hospital-level initiatives directed by international professional medical associations to informal consortia, regional or national governmental bodies to individual NGOs and hospitals.

Source Data

The dataset is sourced from 16 globally recognized medical entities, covering a wide range of healthcare contexts and audiences. For instance, the geographic scope ranges from global (WHO) to national (CDC, NICE) and regional (Ontario, Melbourne) to institutional (ICRC, Mayo Clinic). The corpus also represents health care concerns from high- (Ontario, Melbourne), low- (WHO), and volatile- (ICRC) resource settings. Guidelines also contains a range of technical and conversational vocabulary with target audiences of clinicians or patients (or both), and is sometimes highly specialized within a theme (cancer, pediatrics, infectious disease). The peer review processes also ranged from UN bodies (WHO), institutional review boards (ICRC), professional associations (AAFP) to publicly crowdsourced knowledge bases (WikiDoc). Article length varies widely from very short statements to 100+ page guides.

Data Collection and Processing

PDF documents were converted to text using GROBID. After extracting the raw text from each source, we cleaned data with an ad-hoc process to exclude irrelevant or repetitive content that did not contribute to the textual content, such as URLs, references, figures, table delimiters, and ill-formatted characters. This filtering procedure was performed differently for each source using a sample of 50 articles. Please note that this procedure is not perfect, as it may have removed useful information or kept superfluous content. We provide the raw_text for each article if you would like to perform your own cleaning step. Additionally, the text was standardized to a unified format with hierarchical section headers indicated by '#', homogenous spacing '\n\n' separating paragraphs, and normalized lists formatted with '- ' bullet points. Finally, all samples were deduplicated using title matching, and articles that were too short or not English were filtered out.

Who are the source data producers?

We employed pragmatic selection criteria over medical sources, seeking CPGs that were:

(1) open-access
(2) systematically formatted with homogenous textual structure (i.e., in a format in which automated processes could be deployed without excessive risk of misaligning textual sequences)
(3) in the language predominantly represented by the pre-training corpus of Llama (i.e., English)
(4) covering a breadth of medical sub-domains, audiences (clinician, nurse, patient), and resource settings (high, low, and humanitarian response settings)

Personal and Sensitive Information

As the articles are publicly accessible, no personal or sensitive information is included.

Bias, Risks, and Limitations

Most guideline sources offer reliable and factual information, authored by trusted health professionals. However, users should exercise caution when relying on content from WikiDoc, as it is a crowdsourced encyclopedia. While it generally maintains high quality, there are no guarantees regarding its content.

[More Information Needed]

Recommendations

[More Information Needed]

Citation

To cite the Clinical Guidelines corpus, please use:

@software{meditron2023,
  author = [ADD AUTHORS],
  title = {MediTron-70B: Scaling Medical Pretraining for Large Language Models},
  month = November,
  year = 2023,
  url = {https://github.com/epfLLM/meditron}
}

Authors

Curation: Mary-Anne Hartley
Scraping: Antoine Bonnet, Alexandre Sallinen, Igor Krawczuk
Cleaning: Antoine Bonnet, Alexandre Sallinen