Modeling Emergency Departments Using Discrete-Event Simulation: A Real-Life Case Study Including Patient Boarding

Abstract

This article provides details on modeling and validation of a discrete-event simulation study carried out at emergency department (ED) of a large regional hospital in Belgium. The ED has 21 beds, and a volume of about 30,000 patients per year of which approximately 33% need to be admitted to hospital. Like many other hospital EDs all over world (Pines et al., 2011b), ED we consider in this case study is struggling with a phenomenon called (over)crowding, especially in late afternoon.Following Moskop et al. (2009), we will consistently use term in this article. While there is no single agreed-upon definition of crowding in literature, it can be understood in general as the situation where demand for emergency services exceeds ability of an ED to provide quality care within appropriate time frames (Higginson, 2012). The crowding problem is aggravated by inability of ED to transfer patients that need to be admitted to inpatient wards, due to lack of available inpatient beds. This is referred to as access block or (Crawford et al., 2013; Gilligan et al., 2008; Moskop et al., 2009); by occupying valuable ED space, time, and resources, boarding patients have a negative impact on length-of-stay (LOS) of patients that still require treatment.The model developed in this article reflects patient boarding using time-dependent boarding times and boarding probabilities, which may vary across patient types and are estimated from real-life data. While, in reality, boarding behavior is determined by time-dependent status of beds at inpatient units, this approach avoids a detailed modeling of these units. Although some articles have applied queueing theory (Bekker & Koeleman, 2011; Bretthauer et al., 2011; Cochran & Roche, 2008; Gallivan & Utley, 2011; Koizumi et al., 2005; Lin et al., 2014; Shi, 2013; Thompson et al., 2009) to settings in which both ED and inpatient unit are being considered, it has been recognized in literature that simulation is often preferred tool to study ED operations (Saghafian et al., 2014). The blocking or boarding phenomenon in health care has been studied using simulation from perspective of inpatient wards (for instance; Bagust et al. 1999; Bountourelis et al. 2011; El-Darzi et al. 1998; Mustafee et al. 2012) or with a focus on ED (for instance; Bair et al. 2010; Crawford et al. 2014; Khare et al. 2009; Kolb et al. 2007, 2008; Medeiros et al. 2008; Pines et al. 2011a). None of these studies, however, model ED in much detail. As will be shown, general dynamic behavior in ED is triggered by typical patterns and protocols that have been recognized in literature, and can be acceptably modeled using ED patient record data (thus avoiding detailed data on inpatient units).Section 2 provides an overview of available data, inputs, and assumptions used in simulation model, while Section 3 discusses model validation. Section 4 summarizes main findings. The model was built using Arena R simulation software (V.14) by Rockwell Automation.