Clinical Chatbot Design and Triage Accuracy in Urgent Care Platforms

Abstract

The rapid integration of artificial intelligence in healthcare has positioned clinical chatbots at the forefront of initial patient encounters, particularly within urgent care platforms. This paper systematically investigates the relationship between clinical chatbot design features and diagnostic triage accuracy through clinical modeling based on real-world urgent care platform data. We evaluate how specific user interface designs, conversational structures, and underlying clinical algorithms influence the precision of assigning patient acuity levels. The study utilizes a retrospective observational design, analyzing patient-chatbot interactions and comparing chatbot-generated triage recommendations with physician-validated outcomes. Through extensive clinical pathway analysis, we identify critical design components that mitigate user misinterpretation and enhance data elicitation. Findings reveal that conversational agents incorporating adaptive questioning, empathetic phrasing, and dynamic symptom tracking significantly outperform static symptom checkers in triaging complex urgent care scenarios. The research contributes to the broader health informatics literature by providing empirical evidence on the necessary alignment between user-centered design and clinical risk algorithms. Ultimately, these insights provide actionable guidelines for developing safer, more accurate virtual triage systems in high-volume acute care environments, reducing the risks of both over-triage and under-triage while optimizing resource allocation.