Artificial Intelligence-Based Triage in Emergency Departments: Performance, Workflow Optimization, and Implementation Challenges

Background: Emergency department (ED) triage plays a central role in patient prioritization and resource allocation. However, increasing patient volumes and the growing complexity of clinical presentations have exposed important limitations of traditional triage systems. Artificial intelligence (AI) has emerged as a promising approach to enhance triage through data-driven decision-making. Objective: This study aims to examine the application of AI in ED triage, with a particular focus on predictive performance, workflow efficiency, and implementation challenges. Methods: A narrative review of literature published between 2018 and 2026 was conducted, including systematic reviews, comparative studies, and clinical research. The analysis encompasses machine learning models, natural language processing techniques, and large language models applied to triage tasks. Results: AI-based models demonstrate increased accuracy in predicting clinical outcomes, including hospitalization risk and severity classification. Comparative studies indicate that advanced models—particularly large language models—can achieve performance comparable to clinical decision-making in selected scenarios. In addition, AI-driven tools, such as real-time decision support systems and voice-assisted technologies, contribute to improved workflow efficiency. Nevertheless, several important challenges persist, including data quality issues, model interpretability, bias, and integration into existing healthcare infrastructures. Conclusions: AI demonstrates considerable potential to enhance ED triage by improving predictive accuracy and operational efficiency. However, it should be regarded as a supportive tool rather than a replacement for clinicians. Future research should prioritize scalable architectures, seamless system integration, and validation in real-world clinical environments.