AI Audit Finds Clinical Prediction Models Rarely Share Their Code

AI Audit Finds Clinical Prediction Models Rarely Share Their Code

A review of nearly 4,000 clinical prediction studies finds that reproducible code remains the exception rather than the norm.
LS
Linsey Smith
Oct 9, 2026
1 min read

An AI-assisted review has found that code sharing remains uncommon in clinical prediction-model research, raising questions about how easily published models can be reproduced and independently checked.

Researchers examined 3,967 eligible studies citing the TRIPOD or TRIPOD+AI reporting frameworks. Only 482, or 12.2%, included a code-sharing statement. The rate improved over time, reaching 15.8% among papers published in 2025, but remained far from routine. Studies citing TRIPOD+AI shared code more often than those citing TRIPOD alone: 29.2% versus 11.4%.

The audit also assessed repositories against 14 reproducibility-related features and found wide variation in documentation, dependencies and executable structure. Making code available, in other words, does not automatically make a study reproducible.

The finding echoes the push to open previously restricted scientific AI code and newer attempts to rebuild published studies as interactive, reusable research agents.

The authors argue that clearer expectations for usable, documented code could strengthen clinical prediction research before models move into real-world care.

About the Writer

More from Mindplex

Keep reading

Three more ideas worth your time.

Browse MindBytes

Discussion

Join the discussion

Sign in to share a response with the community.

Type @ to mention someone Type / or use + to add a block Highlight text, then choose Link
Loading editor

Comments cannot be edited after posting because they become part of the reputation record. Give yours a quick review first.