Coupled Graph--Policy Distillation for Personalized Medication Safety in Older Adults with Multimorbidity

2026-08-10

AI Summary

This work addresses medication safety risks faced by older adults with multimorbidity during non-clinical periods due to missed critical health information. It proposes ATLAS, a framework that constructs a medication safety knowledge graph and dynamically generates personalized drug interaction graphs based on patient states. ATLAS incorporates a risk-prioritized multi-agent strategy to perform contraindication screening, risk assessment, and alternative recommendation. The framework innovatively introduces a coupled graph-policy distillation mechanism to translate clinical guideline evidence into individualized decisions and establishes GeriMedBench—the first interactive benchmark for evaluating systems’ evidence-based decision-making capabilities. Experiments demonstrate that ATLAS significantly outperforms existing methods across multiple international benchmarks: on the European non-interactive multimorbidity test, it achieves a 53.73-point gain in Strict Success Rate and a 14.63-point improvement in OSRS over the strongest closed-source LLM baseline, with zero unsafe recommendations in automated evaluation and superior performance across all metrics in clinical double-blind assessment.

Abstract

Large language model (LLM) agents can support medication review between clinical visits, but safe choices for older adults with multimorbidity depend on conditions, medications, and geriatric risks that users may omit. We introduce ATLAS, a coupled graph--policy distillation framework for patient-adaptive medication safety. ATLAS structures guideline evidence as a medication-safety graph. Targeted questions update the patient state and distill relevant relations into a patient-specific medication conflict graph (PMCG). A risk-first multi-agent policy uses the PMCG to screen contraindications, assess cautions and monitoring needs, identify safer alternatives, and verify the final medication plan. We also introduce GeriMedBench, an interactive benchmark that tests safety-critical information acquisition and evidence-based decision revision. Across a European non-interactive multimorbidity benchmark, an Asian interactive multimorbidity benchmark, and an Asian non-interactive cross-guideline benchmark, ATLAS achieves the strongest complete-decision performance among the compared systems. On the European non-interactive multimorbidity benchmark, it exceeds the strongest proprietary LLM baseline by 53.73 points in Strict Success Rate and 14.63 points in overall safety reasoning score (OSRS), with no unsafe recommendations under the automated evaluator. A blinded clinician evaluation gives ATLAS higher mean ratings across all five criteria and flags potentially unsafe recommendations in one ATLAS case and two Gemini cases.

Problem

• medication safety
• multimorbidity
• older adults
• personalized decision-making
• clinical guidelines

Innovation

• graph-policy distillation
• personalized medication safety
• multimorbidity
• medication conflict graph
• risk-first multi-agent policy