TRACER: Time-to-Event EHR Foundation Model
What it is
TRACER is a decoder-based foundation model pretrained on longitudinal electronic health records (EHRs) from over 4.5 million patients across four health systems. It is designed to predict the risk of incident diseases at 1-, 2-, and 5-year horizons, with a specific focus on performance at low false-positive rates suitable for clinical screening.
The problem it solves
Traditional global metrics like AUROC fail to capture the realistic operational constraints of clinical screening, specifically the trade-off between detection and false-alert burden at low false-positive rates.
What you could use it for
- Early detection of incident diseases in patients across different health systems
- Disease screening programs requiring low false-positive rates
- Assessing disease detection and false-alert burden in external, non-development health systems
Why it matters
Enables scalable, early disease detection and screening by leveraging longitudinal EHR data, potentially identifying at-risk patients years before diagnosis while managing false-alert burdens.
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.