Can large language models accurately triage oculoplastic referrals?

Large language models (LLMs) have demonstrated promising performance across a range of medical reasoning and clinical support tasks [ 1 , 2 ]. Accurate outpatient triage is essential to ensure timely access to specialist care while balancing limited health care resources [ 3 ]. Oculoplastic services encompass a broad spectrum of eyelid, lacrimal, and orbital pathologies, ranging from benign disorders to potentially sight- and life-threatening conditions. We evaluated the ability of modern LLMs to assign oculoplastic referral urgency compared with expert triage assessment.