{
  "id": 20812065,
  "title": "AI helps researchers investigate how thousands of medical studies share code",
  "url": "https://urgent.news/2026/11/19/ai-helps-researchers-investigate-how-thousands-of-medical-studies",
  "topic": "science",
  "section": "Science",
  "published": "2026-11-19T22:15:00.000Z",
  "source": {
    "name": "MIT News Research",
    "slug": "mit-news-research",
    "url": "https://news.mit.edu/2026/ai-helps-researchers-investigate-how-medical-studies-share-code-1119"
  },
  "original_language": "en",
  "account": "Sharing data and code from research provides transparency and allows others to examine methods and build upon previous work. Tom Pollard, a researcher at MIT, explains that decisions made during data preparation, model development, and evaluation - \"researcher degrees of freedom\" - can significantly impact study outcomes. While sharing code is a start, it must be well-organized and described to be useful. There is currently little guidance on this, so MIT researchers led an international team to study how code is shared in medical prediction studies. They used large language models to analyze nearly 4,000 papers, identifying statements about shared code and following them to assess hundreds of code repositories. Unfortunately, only about 1 in 8 studies actually shared code. When code was shared, the quality varied greatly - some repositories were well-organized and documented, while others provided almost no guidance. Practices differed greatly between journals and countries. While some journals had low sharing rates but written policies, others had no policies at all. The review also found examples where researchers provided synthetic datasets to assist with code reuse when original patient records couldn't be shared. The team is now developing a reporting guideline called TRIPOD-Code to establish a standard for code sharing in clinical prediction research. This guideline aims to provide practical guidance for researchers and set clearer expectations for journals.",
  "summary": "An MIT and Dana-Farber Institute-led team used large language models to examine nearly 4,000 studies and hundreds of code repositories, revealing gaps in the information needed to scrutinize and reuse published research.",
  "key_points": [
    "Only about 1 in 8 medical studies share their code",
    "Code quality varies greatly between repositories",
    "TRIPOD-Code guideline aims to standardize code sharing"
  ],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}