GLoC-EHR: Evidence-Cited Clinical Reasoning over Global Context and Local EHR Events

2026-10-01

AI Summary

This study addresses the challenge of jointly capturing global trajectories and local events in clinical question answering over structured electronic health records (EHRs) by proposing a multimodal large language model integrated with a dual-memory mechanism. Methodologically, a fixed-size global memory models the overall disease progression while a local memory focuses on critical events. Furthermore, Group Relative Policy Optimization (GRPO) is innovatively introduced to enforce the generation of record-grounded evidence for precise reasoning. Training follows a multi-stage paradigm comprising masked concept pretraining, rationale fine-tuning, and reinforcement learning. Experiments demonstrate that the proposed model achieves state-of-the-art macro-AUROC on MIMIC-IV tasks, significantly reduces ungrounded hallucinations, and exhibits zero-retraining transferability to the EHRSHOT benchmark.

Abstract

Structured electronic health records (EHRs) contain a patient's clinical trajectory as a sequence of clinical codes. Answering clinical questions from such records requires both the context of the whole trajectory and the specific events that support the answer. We introduce GLoC-EHR, a multimodal language model that reads a contextual encoding of the record through a fixed-size global memory of the trajectory and a local memory of selected events. The model learns to generate hospital-course summaries from the global memory and descriptions of masked concepts from the local memory, aligning both with clinical text. It is then trained to cite evidence before answering, through rationale fine-tuning followed by group relative policy optimization (GRPO) with rewards for correct answers and record-supported evidence. On three MIMIC-IV outcome tasks, GLoC-EHR attains the highest macro AUROC among the compared models when it answers directly, whereas zero-shot LLMs reading the serialized record fall far behind. With evidence-cited reasoning, it stays close to its direct multi-task counterpart in macro AUROC, and the evidence terms of the objective reduce unsupported evidence at a similar macro AUROC. The local memory adds distinct supported findings, particularly under strict matching, without a detectable change in macro AUROC. Without retraining, GLoC-EHR transfers to EHRSHOT on par with EHR-BERT and answers two unseen laboratory questions better than zero-shot prompting of its own backbone.

Problem

• Electronic Health Records
• Clinical Question Answering
• Evidence Citation
• Clinical Reasoning
• Multimodal Language Model

Innovation

• Multimodal Language Model
• Dual Memory Architecture
• Evidence-Cited Reasoning
• Group Relative Policy Optimization (GRPO)
• Electronic Health Records (EHR)