Impact of Emergency Department Workload on Triage Behavior
33 Pages Posted: 8 Apr 2022 Last revised: 22 Mar 2023
Date Written: March 31, 2022
Triage is the first step of emergency care, where patients are classified based on their urgency and their anticipated resource usage. Although triage nurses follow a standard classification algorithm, worker judgement plays a significant part in how urgent patients are perceived. In this paper, our goal is twofold: First, we evaluate whether downstream congestion affects prioritization. Specifically, we test whether workload in the treatment area of the emergency department (ED) affects under-triage, defined as perceiving a patient as being less urgent than they truly are. Second, we test the impact of under-triage on patient flow and quality of care. To determine a patient’s true urgency level, and subsequently, define under-triage, we develop a deep-learning model that is trained using information collected during triage. We find that under-triage has a U-shaped relationship with ED workload; it decreases up to the 84th percentile of workload, but increases thereafter. We also find that a one standard deviation increase in under-triage increases patients’ disposition time, room-to-departure times and risk of 30-day readmission by 11.2%, 13.4% and 27%, respectively. From an academic perspective, our work contributes to the healthcare and behavioral operations literature by demonstrating how workload affects customer classification, and quantifying the effects of under-triage. From a practical perspective, our predictive model achieves state-of-the-art performance, and can be employed in hospitals and EDs to assist in patient triage. Also, our results assist managers in making staffing decisions to balance the costs of under-triage.
Funding Information: This research is funded by Boston University, Questrom School of Business research grant.
Declaration of Interests: None.
Keywords: Emergency Department, Triage, Behavior, Workload, Machine Learning, Empirical
Suggested Citation: Suggested Citation