Configurable Clinical Information Extraction with Agentic RAG: What Works, What Breaks, and Why
Researchers have developed an Agentic Clinical Information Extraction (ACIE) system to address limitations in standard retrieval-augmented generation (RAG) for complex patient data. This on-premise RAG pipeline, deployed at University Medicine Essen, is designed to process extensive patient contexts, including temporal reasoning and cross-document dependencies. In a validation study for a lymphoma registry, clinicians verified ACIE's extractions, accepting 96.5% of the data, with acceptance rates varying by extraction type. AI
IMPACT This agentic RAG approach could improve the accuracy and reliability of AI systems in processing complex, heterogeneous clinical data.