Modeling the Emergency Care Delivery System Using a Queueing Approach

Abstract

This thesis considers a regional emergency care delivery system that has a common emergency medical service (EMS) provider and two hospitals, each with a single emergency department (ED) and an inpatient department (ID). Patients arrive at one of the hospital EDs either by ambulance or self-transportation, and we assume that an ambulance patient has preemptive priority over a walk-in patient. Both types of patients can potentially be admitted into the ID or discharged directly from the ED. An admitted patient who cannot access the ID due to the lack of available inpatient beds becomes a boarding patient and blocks an ED server. An ED goes on diversion, e.g., requests the EMS provider to divert incoming ambulances to the neighboring facility, if the total number of its ambulance patients and boarding patients exceeds its capacity (the total number of its servers). The EMS provider will accept the diversion request if the neighboring ED is not on diversion. Both EDs choose its capacity as its diversion threshold and never change the threshold value strategically, and hence they never game. Although the network could be an idealized model of an actual operation, it can be thought of as the simplest network model that is rich enough to reproduce the variety of interactions among different system components. In particular, we aim to highlight the bottleneck effect of inpatient units on ED overcrowding and the network effects resulting from ED diversions. A continuous time Markov chain is introduced for the network model. We show that the chain is irreversible and hence its stationary distribution is difficult to characterize analytically. We identify an alternative solution that builds on queueing decomposition and matrix-analytic methods. We demonstrate through discrete-event simulations the effectiveness of this solution on deriving various performance measures of the original network model. Moreover, by conducting extensive numerical experiments, we provide potential explanations for the overcrowding and delays in a network of hospitals. We suggest remedies from a queueing perspective for the operational challenges facing emergency care delivery systems.