AI helps researchers investigate how thousands of medical studies share code
Summary
- An MIT and Dana-Farber Institute-led team used large language models to examine nearly 4,000 medical studies and hundreds of code repositories.
- The review revealed that code sharing remains uncommon, occurring in only about 1 in 8 studies, and public code often lacks necessary documentation for reuse.
- The findings are informing the development of TRIPOD-Code, a reporting guideline intended to clarify essential requirements for sharing analytical code in clinical prediction research.
Why it matters
Evaluating clinical prediction models requires transparency, but inconsistent and rare code sharing hinders scientific replication. This research leverages LLMs to establish baseline evidence for new reporting standards like TRIPOD-Code.