RECITE uses LLMs and RAG to automate the discovery and extraction of evidence-based revisions for clinical trial eligibility criteria from ClinicalTrials.gov documents. Benchmarked on 3,116 instances, the framework achieves a Cohen's kappa of 0.694 against domain experts in matching original criteria to documented target revisions.
HOW THIS AFFECTS YOU
●
builderYou can leverage this RAG-based approach for specialized document extraction in highly regulated domains.
●
healthThis can streamline the process of broadening clinical trial enrollment through automated, evidence-backed protocol amendments.