Efficacy of an Artificial Intelligence Algorithm for Gatekeeping in Referrals From Primary Care to Specialized Care: a Randomized Controlled Trial

What this Trial Is About

In Rio Grande do Sul, Brazil, the demand for specialty care referrals has greatly increased, especially in rural areas, leading to many automatic authorizations without clinical review. This causes delays for high-risk patients and a large backlog of pending referrals. Researchers are evaluating an artificial intelligence (AI) algorithm designed to triage referrals by urgency and appropriateness to improve the gatekeeping process. This prospective controlled trial randomly assigns referrals to either AI-based triage or standard human review in selected specialty queues. The study compares the standard gatekeeping process, where human evaluators review referrals based on protocols, to an AI algorithm that performs the initial triage of referrals in the electronic system. The AI classifies referrals as authorized or not authorized, and further divides authorized cases into routine or high-risk categories. Referrals with low confidence scores (below 0.8) from the AI are excluded. Subsequent interactions between primary care and gatekeepers may continue to clarify patient needs. Several specialty waitlists will be selected by the health department for the intervention. Participants are referrals from these selected specialties, randomly assigned in a 1:1 ratio to AI or standard gatekeeping. Researchers will measure the proportion of referrals with a final decision within six months as the primary outcome. Secondary measures include time to decision, time to appointment for high-risk patients, use of remote consultations, and waitlist size over time. The study aims to improve referral management and reduce delays in specialty care access, with a total follow-up of six months per referral.

Efficacy of Artificial Intelligence for Gatekeeping in Referrals to Specialized Care