Can Application of Artificial Intelligence Improve Emergency Department Triage Performance?

Emergency department (ED) overcrowding and resulting delayed patient care is a rapidly growing worldwide problem leading to increased mortality and morbidity. Current ED triage systems have a suboptimal ability to differentiate critically ill patients due to subtle signs or symptoms often being difficult to recognize. In recent years, there is growing evidence to suggest that artificial intelligence-based ED-triage systems might enhance an ED clinician's ability to identify patients most in need of intensive ED resources. A systematic review was performed using the American Academy of Academic Medicine Clinical Practice Committee Statement on Protocols for the literature search/grading process to identify articles that could answer the following question: "Can the application of artificial intelligence improve Emergency Department triage performance?"

A structured review was performed of the literature using PubMed. Inclusion criteria were limited to reports from the last 5 years involving human subjects, written in English, and containing the following keywords: Artificial Intelligence, Emergency Department, Triage.

The search yielded 211 results in total, and after the first review, 62 articles were selected to be potentially relevant to our review. The abstract of each article was assessed independently by authors, with 38 selected as responsive for this position statement and reviewed in full.

Based on multiple retrospective and few prospective studies, there is emerging evidence suggesting that Artificial Intelligence-based ED triage systems demonstrate cohort-specific superior performance in predicting adult and pediatric patients who are more likely to need critical care, operative procedures, and hospital admission.