Understanding the Development and Implementation of AI Triage Models in Emergency Departments: A Scoping Review Protocol

Background: Triage is a critical clinical tool in emergency care, ensuring appropriate prioritization of patient care. However, traditional triage systems face challenges such as mistriage, resource misallocation, and overcrowding. Evaluation of triage accuracy also lacks a universal reference standard, making comparisons across studies challenging. Artificial intelligence (AI) has shown potential to enhance triage accuracy, mitigate human limitations, and optimize resource allocation. However, routine integration into clinical workflows remains limited, as uptake depends on staff-related factors such as usability, trust, and organizational readiness. Therefore, there is a need to synthesize evidence on both performance and implementation determinants. Objective: This scoping review compares AI-based triage with conventional/non-AI triage among adults in prehospital and emergency department settings, focusing on implementation determinants as well as performance outcomes. The dual focus on quantitative and qualitative aspects seeks to provide actionable insights for the effective development and integration of AI tools in emergency care. Methods: The review will follow the Joanna Briggs Institute methodology and the PRISMA-ScR reporting guidelines to ensure transparency. We will search PubMed (MEDLINE), Embase (Ovid), Cochrane Central Register of Controlled Trials (Wiley), Web of Science (Clarivate), Scopus (Elsevier), and grey literature via Google Scholar and ProQuest using four core concepts: AI, triage, emergency medicine, and implementation. We will include studies of adult patients comparing AI-based triage with conventional/non-AI triage in prehospital and emergency department settings. Two reviewers will independently screen titles/abstracts and full-text articles and extract data in Covidence®, with a third reviewer resolving disagreements. Quantitative findings will be summarized descriptively in tables/figures, and heterogeneity in triage accuracy definitions and reference standards will be described explicitly rather than pooling non-comparable measures. Qualitative findings will be synthesized thematically and structured using the Consolidated Framework for Implementation Research (CFIR). Discussion: This review will map AI-based triage models and key gaps in design, clinical effectiveness, and implementation across prehospital and in-hospital emergency care. By integrating quantitative and qualitative insights, it will bridge theory and practice, guide future research, and foster human-AI collaboration. Findings may inform the development of more effective AI tools that can be integrated into clinical workflows and strengthen acute patient care.