LLMs as Clinical Research Assistants: Secure and Accurate Extraction from Unstructured EHR Narratives

This paper presents a secure, modular framework that leverages locally deployed large language models (LLMs) to automate structured feature extraction from unstructured electronic health record (EHR) narratives. The system integrates retrieval augmented generation (RAG) and structured response methods to reduce the burden of manual chart review, increase consistency in data capture, and accelerate clinical research, achieving high accuracy across multiple medical characteristics.