AI-assisted scoping review of code sharing in clinical prediction model research.

Only 12.2% of 3,967 clinical prediction model articles include code-sharing statements, with significant variation across journals and countries.

• Why it matters: Limited code sharing hampers reproducibility and independent validation of clinical prediction models, which are crucial for reliable diagnostic and prognostic decision-making.
• What they did: A large-language-model-assisted pipeline analyzed 3,967 articles citing TRIPOD or TRIPOD+AI, extracting repository links and evaluating 482 repositories against 14 reproducibility features.
• The result: Findings reveal substantial heterogeneity in reproducibility practices, highlighting the need for clearer guidelines on documentation, dependency management, and executable structures to enhance usability.