About 80% of trials miss their enrollment targets, and 11% of activated sites never enroll a single patient. The workflow bottleneck is fundamentally manual: coordinators reviewing fragmented EHRs against 20 to 40 protocol criteria, often missing eligible patients simply due to volume.
Recent evidence shows AI-assisted eligibility screening can help. A 2024 study covering 3,281 Phase 2 and 3 trials used NLP to convert free-text eligibility criteria into structured, machine-readable conditions matched against real patient data. A pilot at the Medical University of South Carolina tested automated EHR surveillance across five cardiovascular and cancer trials, achieving recall up to 0.778 without disrupting clinical workflow. In ophthalmology, a deep learning model identified 64% more eligible patients than keyword search, with higher imaging-criterion precision.
A 2024 scoping review of 51 studies found AI recruitment tools improved efficiency, reduced costs, and increased accuracy, but noted small study populations, varied outcome measures, and fairness concerns in 11 studies. Implementation barriers include data interoperability limits, algorithmic bias risks, and regulatory uncertainty. AI should accelerate candidate identification, not replace investigator judgment or informed consent workflows.