Is ESI Dead? Examining the Potential for AI to Transform Emergency Department Triage

The Emergency Severity Index (ESI) has served as the national standard for more than twenty years, providing an effective framework for millions of emergency department ED visits each year. As patient volumes continue to rise and clinical complexity grows, the field has increasingly explored whether modern data-driven methods can support safer and more accurate triage decisions. Over the last several years, multiple lines of research have shown the promise of machine learning and AI in this domain. Studies leveraging structured EHR data, vital signs, utilization patterns, and clinical context have demonstrated improved prediction of short-term return visits, critical illness, and hospital admission. More recently, real-world deployments of AI-informed triage recommendations have shown shifts in acuity distribution, more precise identification of high-risk patients, and measurable improvements in patient flow.  This State-of-the-Art session will bring together leaders and researchers in triage science, AI development to examine the future of triage in EM. We will synthesize the existing literature, discuss current vendor-based solutions, and present data from our own large-scale AI triage research to demonstrate how next-generation tools could be safely and transparently integrated into clinical workflows. We will explore what a hybrid human–AI triage systems and what this evolution means for emergency medicine practice, operational efficiency, and patient outcomes. This SOTA will highlight preliminary research findings that leveraged nearly 1,000,000 ED triage encounters from the Mount Sinai Health system to examine how an AI-augmented system that can impact ESI and prediction of deterioration.