The use of artificial intelligence in emergency department triage: Evaluation of patient routing accuracy

The Use of Artificial Intelligence in Emergency Department Triage: Evaluation of Patient Routing Accuracy Aim: This study aims to evaluate the accuracy and effectiveness of using artificial intelligence (AI) in the triage of patients admitted to the emergency department at Pamukkale University Hospital. Materials and Methods: The research was conducted between June 2024 and November 2024, including a total of 788 patients. Participants' complaints, vital signs, medical histories, and brief anamneses were recorded and processed through an AI software based on ChatGPT-4o. Triage assessments were performed by emergency medical technicians (EMTs) and subsequently verified by an emergency medicine specialist. The data were analyzed using the Kolmogorov-Smirnov test, Kappa test, and Bland-Altman method. Findings: The average age of the 788 patients included in the study was 42.02 ± 18.66 years. The gender distribution was 395 females (50.1%) and 393 males (49.9%). The distribution of patients by triage categories revealed that the AI-based system had an accuracy rate of 92.41% in the red category, 84.62% in the yellow category, and 95.21% in the green category. In comparison, EMTs had an accuracy rate of 43.04% in the red category, 91.35% in the yellow category, and 75.82% in the green category. Age group analysis showed that AI performed particularly well in the elderly patient group, while EMTs had relatively lower performance in the young adult group. When analyzed by gender, it was found that AI had higher accuracy rates for both male and female patients. When patient complaints were examined, it was found that AI had significantly higher accuracy rates compared to EMTs, especially in high-risk cases such as chest pain and shortness of breath. Conclusion: The applicability and effectiveness of artificial intelligence technologies in emergency department triage have shown positive outcomes. AI-assisted systems xiii have the potential to improve decision-making processes in triage, significantly enhancing patient safety and optimizing resource utilization in emergency departments. This research provides a significant contribution to the literature on the effectiveness of AI in the field of emergency medicine and serves as a basis for future studies. Keywords: Emergency Department, triage, artificial intelligence, Chatgpt-4o, health technologies, automated triage systems, medical decision support systems