Enhancing Clinical Trial Patient Matching through Knowledge Augmentation and Reasoning with Multi-Agents

Abstract: Matching patients effectively and efficiently for clinical trials is a significant challenge due to the complexity and variability of patient profiles and trial criteria. This paper introduces Multi-Agents for Knowledge Augmentation and Reasoning (MAKAR), a novel framework that enhances patient-trial matching by integrating domain-specific knowledge with structured reasoning. MAKAR consists of two key modules: the Augmentation Module and the Reasoning Module. The Augmentation Module enriches trial criteria by incorporating detailed explanations of relevant concepts to ensure clarity and completeness. The Reasoning Module then evaluates each health condition, following a structured, step-wise approach to determine eligibility and make the final matching decision. This paper enhances patient-trial matching by leveraging the agency and reasoning capabilities of LLMs through automated agent interactions, including collaboration, critique, and navigation. Experimental results on a public dataset demonstrate that our framework surpasses existing benchmarks, achieving up to an 8% improvement in accuracy for specific criteria. Furthermore, in a real-world offline test, MAKAR achieved a 100% accuracy. These findings show MAKAR's potential as a scalable and robust solution for clinical trial patient matching.