LLM summarizes patient notes to improve hospital outcome predictions
RASPER: Reward-Aligned Summarization of Clinical Notes for EHR Outcome Prediction
Artificial Intelligence
Summary
Clinical records have lots of detailed notes that are hard for computers to understand but contain important patient information. The authors created RASPER, a tool that shortens these notes while focusing on details useful for predicting health outcomes like readmission or medication needs. It learns to keep the most relevant facts by using feedback from prediction tests instead of just making the summary sound good. This helps hospitals use both codes and notes together to make better decisions about patient care.
What this means in practice
- •For hospital data teams: Integrate improved note summarization to enhance predictive models for patient readmission and medication needs.
- •For medical software developers: Develop clinical decision support tools that use reward-aligned summaries to provide more accurate patient outcome predictions.
Authors
Arya Hadizadeh Moghaddam, Mohsen Nayebi Kerdabadi, Chen Chen, Dongjie Wang, Zijun Yao
Abstract
Unstructured discharge notes in Electronic Health Records (EHRs) often carry signal complementary to structured medical codes, holding patient-specific evidence that standardized cohort-level codes alone cannot capture. However, this evidence in notes is frequently buried in lengthy, noisy text that is not intentionally written with any specific clinical prediction in mind. Summarization is an obvious mitigation, but generic summaries, tuned for fluency rather than the outcome, routinely omit decisive evidence while retaining plausible but uninformative detail. To this end, we propose RASPER, a Reward-Aligned Summarizer for Prediction in EHR, that optimizes note summarization directly against the downstream clinical task. RASPER employs a tunable LLM-based summarizer to extract task-relevant evidence from discharge notes and trains it via reinforcement learning from prediction feedback, using a reward derived from the downstream predictor's loss. To ground the summarizer, a longitudinal encoder converts structured codes into soft prompts that incorporate each patient's clinical context into note summarization. By rewarding the quality of the resulting multimodal prediction, RASPER encourages the summarizer to retain patient-specific evidence that complements, rather than duplicates, information captured by structured codes. RASPER consistently outperforms strong baselines on both readmission prediction and medication recommendation across MIMIC-III and MIMIC-IV.