The abstract is read at the publisher; the summary is JClub's.
MENTOR selected institution-specific negative control outcomes in a Korean tertiary hospital EHR, proposing 243 NCOs with higher accuracy 0.83 than CEM-only 0.53.
- Why it matters: Negative control outcomes are used to detect systematic error in observational studies, but choosing plausible, institution-specific true nulls is still largely manual and time-consuming.
- What they did: Four LLM agents used the Observational Medical Outcomes Partnership common data model, combining Common Evidence Model candidate lists with LLM reasoning and a local patient-count relaxation to reach a target number of NCOs.
- The result: Compared with CEM-only and first-pass LLM-only selection across 349 candidates for three drug pairs, MENTOR improved sensitivity (0.89 vs. 0.41) and PPV (0.86 vs. 0.79), with judge-expert agreement kappa=0.63.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.