# RELEAP: reinforcement-enhanced label-efficient active phenotyping for electronic health records

## TL;DR

RELEAP improves EHR-based risk prediction by using reinforcement learning to guide phenotype correction and sample selection under labeling budget constraints.

## Contribution

RELEAP introduces a reinforcement learning framework that directly optimizes downstream prediction performance for active phenotyping.

## Key findings

- RELEAP improved logistic AUC from 0.774 to 0.807 and survival concordance index from 0.715 to 0.749.
- Performance gains were stable across iterations and consistent in sex-stratified subgroup analyses.
- RELEAP outperformed proxy-only baselines and approached oracle performance under the same labeling budget.

## Abstract

Electronic health record (EHR) phenotyping often relies on noisy proxy labels, which undermine the reliability of downstream risk prediction. Active learning can reduce annotation costs, but typical heuristics do not directly optimize downstream prediction. Our goal was to develop a framework that directly uses downstream prediction performance as feedback to guide phenotype correction and sample selection under constrained labeling budgets.

We propose reinforcement-enhanced label-efficient active phenotyping (RELEAP), a reinforcement learning-based active learning framework. Reinforcement-enhanced label-efficient adaptively integrates multiple querying strategies and, unlike prior methods, updates its policy based on feedback from downstream models. We evaluated RELEAP on a de-identified Duke University Health System (DUHS) cohort (2014-2024) for incident lung cancer risk prediction, using logistic regression and penalized Cox survival models. Performance was benchmarked against noisy-label baselines and single-strategy active learning.

Reinforcement-enhanced label-efficient improved over the proxy-only baseline and approached oracle performance under the same budget. Logistic AUC increased from 0.774 to 0.807. Survival concordance index increased from 0.715 to 0.749. Gains were stable across iterations using downstream feedback. These trends were consistent in sex-stratified subgroup analyses (female vs male).

By linking phenotype refinement to prediction outcomes, RELEAP learns which samples most improve downstream discrimination and calibration, offering a more principled alternative to fixed active learning rules.

Reinforcement-enhanced label-efficient optimizes phenotype correction through downstream feedback, offering a scalable, label-efficient paradigm that reduces manual chart review and enhances the reliability of EHR-based risk prediction.