Clinical Trial Patient Finder LLM

Executive summary

Patient recruitment for clinical trials is a persistent bottleneck in clinical research, often leading to delayed timelines, increased costs and underpowered studies. Matching patients to trials requires interpreting complex, free-text eligibility criteria and comparing them against structured but heterogeneous patient data, a process that is traditionally manual, slow and error-prone. As a result, many industry-sponsored trials struggle to meet enrollment targets, limiting the pace of medical innovation.

This piece demonstrates a practical application of Retrieval-Augmented Generation (RAG) using large language models (LLMs) to improve patient-to-trial matching workflows. Rather than focusing on chatbot interactions, this system showcases how RAG can be repurposed to interpret unstructured eligibility text and dynamically filter patient datasets based on those criteria. By automating eligibility translation and matching logic, the approach highlights how LLMs can reduce manual effort, improve consistency and support scalable clinical trial recruitment systems.