Some machine’s doin’ that for you*—electronic triage systems in the emergency department

Abstract

Background: In the last 25 years, triage systems for urgency assessment have been established in emergency departments (ED). The existing symptom-based triage systems show some weaknesses. The present state of digitization of the EDs and the capacity of the current generation of computers allow, at the time of triage, algorithm-based data evaluation and risk stratification for specific clinical endpoints beyond the triage level. Aim of the study: After selective literature review, we performed an analysis of electronic triage systems (ETS). The underlying principles and the current capabilities of ETS are explained. In addition, opportunities and difficulties of establishing ETS in German EDs are considered. Results: Age and other vital signs have been identified as important predictor variables, which have not yet been standardized in triage systems. Using these variables, machine learning (ML) can provide reliable predictions for clinical endpoints such as inpatient admission or mortality. The overall quality of initial assessment via ETS is high. ETS can offer evidence-based support to triage staff in patient disposition and reduce over- and undertriage. There are developments that create favorable conditions for the use of ETS in German EDs. For example, representative ED registries facilitate the creation of reference data sets needed to build computer-based classification models. In addition, individual patient data must be available quickly. Conclusion: ETS can help increase patient safety and improve resource utilization. However, to date, objective reference standards and guidelines for machine learning are still lacking.