
AI-Assisted Scoping Review of Code Sharing in Clinical Prediction Model Research
AI-assisted scoping review of code sharing in clinical prediction model research
Research Summary
Study Title: AI-assisted scoping review of code sharing in clinical prediction model research
Publication:
Dana-Farber Cancer Institute authors: Thomas Sounack, Charlotta Lindvall, MD, PhD
Summary: Dana-Farber Cancer Institute researchers in collaboration with colleagues at MIT are publishing results of a new study in Nature Medicine which examined the availability and reproducibility of code underlying clinical prediction models. Using an AI-assisted approach, researchers reviewed approximately 4,000 published studies and found that code was shared in only 12% of studies—and even when code was shared, it was rarely reproducible. Sharing prevalence was also found to vary widely by journal and country.
Significance: The study findings have significant implications for the use of AI and predictive models in healthcare. Without reproducible code, researchers cannot independently verify published results, making it more difficult for clinicians, regulators, and health systems to evaluate whether models are reliable enough for clinical use and patient care. Strengthening these practices in clinical AI research may improve the usability of clinical prediction model studies and support their deployment in real-world clinical settings.
Funding: Funding was primarily provided through the National Institutes of Health
Related to this article








