ClinNoteAgents: An LLM Multi-Agent System for Predicting and Interpreting Heart Failure 30-Day Readmission from Clinical Notes

2025-12-07

AI Summary

High hospital readmission rates for heart failure (HF) remain a critical clinical challenge, yet rich risk information embedded in unstructured clinical notes is underutilized. This paper proposes a large language model (LLM)-driven multi-agent system that automatically extracts clinical and social risk factors from free-text notes—without requiring extensive labeled data or predefined structured fields—and generates interpretable feature representations to support 30-day readmission risk prediction. The system integrates information extraction, semantic parsing, and causal reasoning to enable risk factor association analysis and generate physician-style decision summaries. Evaluated on 2,065 real-world patient records, the method significantly improves risk factor identification accuracy and predictive performance (AUC increased by 4.2 percentage points), while ensuring interpretability, scalability, and clinical utility.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Multiagent Systems: Modeling other Agents

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Large language models for search

Abstract

Heart failure (HF) is one of the leading causes of rehospitalization among older adults in the United States. Although clinical notes contain rich, detailed patient information and make up a large portion of electronic health records (EHRs), they remain underutilized for HF readmission risk analysis. Traditional computational models for HF readmission often rely on expert-crafted rules, medical thesauri, and ontologies to interpret clinical notes, which are typically written under time pressure and may contain misspellings, abbreviations, and domain-specific jargon. We present ClinNoteAgents, an LLM-based multi-agent framework that transforms free-text clinical notes into (1) structured representations of clinical and social risk factors for association analysis and (2) clinician-style abstractions for HF 30-day readmission prediction. We evaluate ClinNoteAgents on 3,544 notes from 2,065 patients (readmission rate=35.16%), demonstrating strong performance in extracting risk factors from free-text, identifying key contributing factors, and predicting readmission risk. By reducing reliance on structured fields and minimizing manual annotation and model training, ClinNoteAgents provides a scalable and interpretable approach to note-based HF readmission risk modeling in data-limited healthcare systems.

Problem

• Predicts heart failure 30-day readmission risk from clinical notes
• Extracts structured risk factors and clinician-style abstractions automatically
• Reduces reliance on manual annotation and structured EHR data

Innovation

• LLM multi-agent system extracts structured risk factors from clinical notes
• Transforms free-text notes into clinician-style abstractions for prediction
• Reduces reliance on structured fields and manual annotation for scalability