Few-Shot Large Language Models for Actionable Triage Categorization of Online Patient Inquiries

ArXi:2605.15680v1 Announce Type: cross Online patient inquiries are often informal, incomplete, and written before professional assessment, yet they must still be routed to an appropriate level of clinical follow-up. We study this as a four-class actionable triage task -- self-care, schedule-visit, urgent-clinician-review, or emergency-referral, and ask whether prompted large language models (LLMs) can such routing under low-resource labeling conditions.