Abstract
Emergency departments (EDs) worldwide face increasing pressure to optimize triage processes amidst rising patient volumes and resource constraints. Artificial intelligence (AI) has emerged as a potential solution to enhance triage accuracy and efficiency, yet its real-world clinical impact remains inadequately characterized. We conducted a systematic review following Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines, searching PubMed/Medical Literature Analysis and Retrieval System Online (MEDLINE), Excerpta Medica Database (Embase), Web of Science, and Institute of Electrical and Electronics Engineers (IEEE) Xplore (2020-2025) for studies evaluating AI-based ED triage systems. From 119 initially identified records, six studies met inclusion criteria after duplicate removal (n=67), title/abstract screening (n=52), and full-text assessment (n=12). Eligible studies reported quantitative outcomes on AI performance compared to traditional triage methods. Risk of bias was assessed using an adapted Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool. Narrative synthesis was employed due to methodological heterogeneity. The included studies (n=6) demonstrated AI's potential to reduce triage time, improve documentation accuracy, and enhance decision support. Voice-based artificial intelligence (Voice-AI) systems achieved 19% faster documentation versus manual methods, while machine learning algorithms reduced mis-triage rates by 0.3-8.9%. However, limitations included undertriage risks, variable accuracy, and predominance of single-center studies. Implementation challenges encompassed workflow integration barriers and insufficient clinician acceptance metrics. AI-based triage systems show promise for improving ED efficiency but require rigorous multi-center validation and standardized outcome reporting. Key gaps include evidence on patient-centered outcomes, equity considerations, and long-term impact studies. Future development should prioritize explainable algorithms, clinician engagement, and ethical frameworks to ensure safe implementation.