27 Expanding Your Operational Toolkit: Using Computer Simulation to Improve Flow

Abstract

Emergency department (ED) crowding has become a crisis for the U.S. health care system. Crowding often manifests as long waiting times for individual patients, extended overall length of stay (LOS), and high left-without-being seen (LWBS) rates. Hospitals and health care systems must invest in management strategies to address ED overcrowding and inpatient boarding. Computer simulation may be a cost-effective, efficient means to proactively identify and manage bottlenecks in patient care and excessive wait times or prolonged LOS. Simulation models are a novel tool to enhance prospective operational planning. We aimed to develop and validate a simulation model to estimate LOS in our ED. We then aimed to evaluate whether the simulation model could be used for operational planning by examining whether it could accurately predict the impact of planned operational changes. A high-fidelity, highly granular discrete event simulation (DES) model of patient flow through our ED was created in MedModel, a health care modeling operations software from the ProModel Corporation (Allentown, PA) that relies on queuing theory. This in silico model includes the entire ED, allowing visualization and analysis of multiple stages/queues during the patient flow through the ED. The model was developed using time stamps abstracted from the electronic health record (EHR): bed placement, RN assessment, MD assessment, and turnaround times (TOTs) for laboratory or imaging studies. Initial validation was performed by comparing system output LOS times to actual LOS values. The model was considered to be a valid predictor of LOS if simulated estimates were within a 5% margin of the actual LOS. After initial validation, the simulation was asked to model the impact of a dedicated inpatient holding unit. In the real world, a 15-bed ED pod was converted into a holding unit for admitted patients boarding in the ED awaiting inpatient bed assignment. LOS metrics were compared between simulation projections and actual data after implementation of this change. In this case, the model was considered valid if simulated estimates were within a 5% margin of the actual LOS and in the same direction of action. Initial validation of our simulation model confirmed that the model would predict LOS within 5% of real-world metrics. Simulated LOS was accurate, predicting LOS that was within 4.1% of actual LOS for all patients, and 0.8% of LOS for admitted patients. The simulation model also accurately predicted direction of impact on LOS for admitted patients following implementation of a 15-bed inpatient holding unit in the ED (Table 1). A DES simulation model is a valid tool to estimate LOS in the ED and can predict LOS following operational interventions (eg, implementation of an inpatient holding unit). Future work should test whether or not use of DES as an adjunct to standard decisionmaking improves operational metrics.Table 1Simulation model performance evaluating creation of inpatient holding unit.SimulationActual% DifferenceAdmit LOS, pre-implementation721 min726.4 min0.7%Admit LOS, post-implementation704 min715.8 min1.6% Open table in a new tab