HEAL: Healthcare Emergency Assistants Leveraging Large Language Models

Emergency Department (ED) is an essential component of health-care systems, providing 24-hour care to patients who require urgent medical treatment and attention. Over time, ED activity has increased rapidly while the number of staff has often remained constrained, consequently leading to overcrowding and potentially compromised patient care. The emergence of advanced Large Language Models (LLMs) presents a unique opportunity to address such a challenge. In this paper, we introduce an innovative system, HEAL, to automate and optimize the conversations and information extraction components involved in the triage process in EDs. HEAL operates in two distinct modes: i) Registration & Triage Mode for collecting initial patient information and categorizing triage levels, and ii) Monitoring Mode for continuous evaluation of conditions of patients in waiting rooms. Furthermore, we demonstrate the effectiveness of our system in patient information collection and triage categorization on our collected dataset. Video illustration and more materials can be found in https://github.com/Zihan-Zhu/HEAL.