Reproduce “TRACER, a time-to-event EHR foundation model for disease detection across health systems” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Mara Pleasure, Ekaterina Redekop, Vedrana Ivezic, Rushikesh Kulkarni, Aaron Chin, William Speier, Corey W. Arnold
TRACER estimates future disease risk from longitudinal EHR histories and evaluates screening thresholds across diseases, horizons, and held-out health systems. The model was pretrained on more than 4.5 million patients and tested on 16 diseases. At one year and 95% specificity, heart-failure and pancreatic-cancer PPVs were 13.80% and 0.71%; these retrospective estimates do not establish patient benefit.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.