This paper presents a structured approach to building a high-fidelity simulation for an emergency department. Our approach has three key features. First, we use the concept of modules as a building block for modeling. A module is a minimum unit that has clinical or administrative meanings in ED operation, and it consists of low level operational activities. Second, we use a structured template to formally represent modules, and we adopt notations and grammars from the business process modeling notation. This provides an enhanced clarity and transparency, which proves very useful in extracting necessary data from a hospital database or from interviewing ED staff. Finally, we define an interface, specifically data structure and handler, for converting information represented in the modules into simulation languages. This interface makes it possible to seamlessly link the modeling process to the implementation process in the simulation construction.
Improving an emergency department (ED) for better patient care requires solutions for a wide range of problems. Such solutions typically involve human resource scheduling (Rossetti, Trzcinski and Syverud 1999; EL-Rifai et al. 2015), bed capacity adjustment (Ahmed and Alkhamis. 2009; Zeinali, Mahootchi and Sepehri 2015), and streamlining discharge process (Khare et al. 2009; Powell et al. 2012; Shi, Dai and Ding 2015). While solution strategies in general are well known, designing and implementing specifics of these solutions is never straightforward due to the complexity of operational environment of ED. Trial-and-error is not a desirable approach since introducing changes in an operating ED can be quite costly(Jun et al. 1999); introducing changes that turn out ineffective could cost resources invested for the changes, but more than that, it could mean loss of opportunities for saving lives. Thus, it is critical to have reasonable prediction on the effects of a solution to be implemented before actually implementing it.
Simulation is a natural tool to answer to those what-if questions, as evidenced by many prior literature on ED simulation studies. Paul, Reddy and DeFlitch (2010) offer an excellent surveys on the studies that use simulation models for solving problems in EDs. Simulation models offer a great deal of flexibility and rich details so that complexity of an ED can be captured. By conducting simulation experiments, we can test many things that cannot be easily tested in a real ED without incurring cost or risks accompanying real-world experiments. Of course, a fundamental premise is that a simulation model that we use for such experiments behaves as close to the real-world ED as possible; it has to be a precise replica of the ED that we intend to study (Sargent 2005). We must construct a high-fidelity simulation model of a target ED system. While specific requirements for a high-fidelity ED simulation can vary, it may be characterized by detailed descriptions of actual ED operations such as immense diversity in patient pathways, realistic resource activities, spatial trajectories of patients and caregivers, and detailed level of care processes, etc.
To construct a high fidelity simulation is a challenging task both in terms of modeling and implementation into a simulation program. First of all, it is difficult to extract sufficient and relevant information on a target ED system. Communication between simulation modelers and ED staff can be ineffective and inefficient due to differences in their background as well as the nature of specialized knowledge. In addition to the commitment from higher-level management, which is often cited in the ED simulation literature as a critical success factor for a simulation project, a systemic framework is needed to more effectively engage all parties involved. Second, unstructured information obtained from surveys and interviews often contain large amount of information irrelevant to simulation modeling. Even information from an electronic medical record (EMR) system requires substantial efforts for data cleaning and deciphering. Data from an ED EMR system does not simply produce a nice and clean time-stamped patient flow information because they are not developed to serve such purpose. It is most likely that we need to combine data from an EMR system and the estimates and judgment from ED staff. This problem becomes more severe as we incorporate higher level of details in a simulation model. Third, in its implementation phase, we need an efficient means to transfer large amount of information captured in a reference model to a simulation model.
Sample implementation in AnyLogic 7.2.0.
Based on our most recent experience of building an ED simulation model, this paper introduces a structured approach for constructing high fidelity ED simulation. Our approach has three key features. First, we use the concept of modules as a building block for modeling. A module is a minimum unit that has clinical or administrative meanings in ED operation, and it consists of low level operational activities. Since a module can be understood in the clinical or operational context, communication is made much easier and clearer. Second, we use a structured template to formally represent modules, and we adopt notations and grammars from the business process modeling notation (BPMN). The formal representation include specifications for which information will be obtained from ED EMR system or staff knowledge. This provides an enhanced clarity and transparency, which proves very useful in extracting necessary data from a hospital database or from interviewing ED staff. Finally, we define an interface, specifically data structure and handler, for converting information represented in the modules into simulation languages. This interface makes it possible to seamlessly link the modeling process to the implementation process in the simulation construction.