Artificial Intelligence-Based Clinical Decision Support in the Emergency Department: Bridging Development to Implementation

Abstract

Background: Artificial intelligence (AI)-based clinical decision support (CDS) tools are desired by physicians to augment high-stakes decision-making in the emergency department (ED). Objective: This thesis evaluates the current field of AI-CDS in the ED and explores barriers and facilitators to their development and implementation. Methods: We conducted a scoping review of AI-CDS tools for individual ED patient care and categorized them by phase of development. We conducted interviews with expert researchers to identify barriers and potential facilitators for successful implementation. Results: Despite a rapidly growing number of publications, only 3.5% of AI-CDS tools have been tested or implemented in a live clinical setting. Expert researchers identified challenges regarding data infrastructure, team capacity, defining the clinical problem, regulatory approval, legal and liability concerns, time, and cost. Conclusion: To bridge the gap between development and implementation, researchers must incorporate implementation science principles in the earliest stages of AI-CDS tool development.