AI-Assisted Clinical Trial Patient Recruitment

Abstract

Clinical trial recruitment remains slow because eligibility criteria are complex, patient information is fragmented, and screening relies heavily on manual review. Recent artificial intelligence systems can retrieve trials and classify eligibility, but many treat patient suitability as a fixed binary decision. This study addresses the lack of recruitment models that account for temporal clinical changes, missing evidence, fairness, and uncertainty simultaneously. It proposes a human-supervised, uncertainty-aware framework that continuously evaluates structured records, clinical narratives, laboratory trends, and trial criteria. The framework distinguishes confirmed eligibility, confirmed ineligibility, possible eligibility, insufficient evidence, and time-dependent eligibility. A fairness-sensitive ranking mechanism is introduced to prevent high-volume demographic groups from dominating candidate recommendations. The proposed approach aims to reduce unnecessary chart review while preserving clinician control, evidential traceability, and patient privacy. The study establishes a foundation for safer and more inclusive AI-assisted recruitment across multiple clinical specialties