The Cost of Task Switching: Evidence from Emergency Departments

Abstract

Problem Definition: Emergency department (ED) physicians treat patients with different symptoms and constantly switch between tasks. Utilizing three years of comprehensive data on patient visits and lab tests from two large EDs, we investigate the impact of task switching on physician productivity, quality of care, and patient routing. We subsequently provide operational solutions based on the identification findings.

Methodology: To address estimation bias due to measurement errors and endogenous patient selection, we refine the sample period and construct an instrumental variable called switching likelihood, which exploits the exogenous composition of waiting patients. By exploring the heterogeneous impact on physician productivity among different patient type pairs from data, we leverage a max bisection algorithm to partition patients into two clusters to minimize the negative impact of task switching.

Results: Our estimates indicate that, at different EDs, a 10% increase in the switching frequency of physicians reduces the number of patients treated per hour by 8.65% - 11.53% on average. However, we find no significant influence on treatment quality. We propose a data-driven queue management method to optimally partition patients into two queues. Based on the simulation of implementing the proposed two-queue system in our collaborating EDs, we find that the average waiting time is reduced by up to 40%.

Managerial Implications: Task switching negatively impacts ED physician productivity, and this impact is more prominent for certain patient type pairs. Being aware of the switch cost, we propose measures to mitigate switch costs, which can considerably reduce ED congestion and patient waiting times.

Keywords: Task switching, emergency department, behavioral queueing, data-driven, queue management, empirical, max bisection, graph