Emergency departments (EDs) are under increasing pressure due to growing patient volumes, limited resources, and long waiting times. Accurate and efficient triage is essential for patient safety, but traditional systems like RETTS can vary in reliability, depending on the nurse’s assessment. This study explores whether an AI-based triage system can help address these challenges by supporting clinical decision-making.

A prototype AI triage tool was developed using a large language model to simulate patient interactions and assign triage levels based on self-reported symptoms.The system was evaluated through 15 simulated cases and compared to assessments made by experienced triage nurses. The study combined a comparison of triage decisions with input and reflections from the nurses to assess the tool’s performance and usefulness.

The results showed that the AI matched the nurses’ triage decisions in 10 out of 15 cases (66.7%). In the remaining cases, the AI tended to assign a higher urgency level, reflecting a safety-first approach. Nurses appreciated the tool’s structure and speed but noted limitations due to the lack of vital signs and non-verbal cues.

This research suggests that AI-based triage tools have the potential to improve efficiency and patient prioritization in EDs. While the system cannot replace human judgment, it could serve as a valuable support tool, especially during busy periods. Further development and real-world testing are needed to assess how such systems could be integrated into everyday healthcare workflows.