Increasing demand for healthcare services, due to changes in demographic shifts and constraints in healthcare funding, make it harder to manage effective, sustainable healthcare systems. Many healthcare modeling exercises have been undertaken with the aim of supporting the decision-making process. This paper reviews all of the 456 articles published by the Winter Simulation Conference over the past 48 years (1967–2015) on the subject of modeling and healthcare system simulation, and analyzes the relative frequency of approaches used. A multi-dimensional taxonomy is applied to encompass the modeling techniques, problem applications and decision levels reported in the articles. One of the most significant changes in the modeling of healthcare systems is the fact that Discrete-event Simulation (DES) is no longer used as an autonomous method, but rather as an integral part of the solution. The , hybrid and multi-paradigm approaches feature strongly in the current direction of modeling in healthcare systems.
Health systems can be described as complex systems characterized by a high level of uncertainty and dynamism (Rashwan et al. 2013), as well as much variability (Brailsford 2007). Variability exists in all systems and can have an enormous impact on productivity and performance. In the healthcare context, variability results from factors within our control (such as staff training, checklist, room temperature) and beyond our immediate control (such as unscheduled patients, treatment time under complex conditions). Variability can be disruptive in any system, to various degrees. Therefore, the need for techniques and sophisticated tools that make it possible to measure, understand and manage variability will always top the wish-list of management teams.
Operations Research (OR) has contributed strongly to understanding the different levels of complexity of healthcare processes, including the variability and uncertainty of activities. Over the past 50 years, OR researchers have worked closely with professions in the management of the healthcare system, seeking to offer a diverse portfolio of solutions to address the current issues at different decision levels. Modeling the unit, process or system has always been the first phase of most of the studies, regardless of the algorithm or framework applied as a solution.
One of the important factors to ensure a good model is the quality of the data phase. The modeling outcome is dependent on the data from different levels, particularly when the simulation constitutes the ultimate environment for experiments. Every data level can provide insights that can be used to understand the system. For example, level 0 data identifies a portion of the real-world system that will be modeled. As the level is higher, the measurement and observations make more sense. Level 3 consists of the structured data, which, along with the modular data, is used for the modeling. The complexity of the system depends on the availability and accuracy of the data levels (Zeigler, Praehofer, and Kim 2000). The modeling of healthcare systems often suffers from a dearth of data, especially at level 0 and 1.
Several review papers have been published over the years on healthcare modeling and simulation. Jun, Jacobson, and Swisher (1999) reviewed the modeling of healthcare systems, with an emphasis on Discrete-event Simulation, while Fone et al. (2003) conducted a systematic review of the use of simulation modeling in population health. Fakhimi and Mustafee (2012) focused on UK healthcare modeling. A comprehensive review, which examined 342 articles based on nine criteria, was conducted by (Brailsford et al. 2009). Their review describes a multi-dimensional method for classifying the literature on simulation and modeling in the healthcare context. This paper reviews the publications of the Proceedings of the Winter Simulation Conferences (WSC) (1967–2015).
Mixing methods can compensate for the weaknesses and drawbacks of a single method (Brailsford 2015). Numerous studies presented during WSC proceedings have attempted to combine simulation methods with other methods. Over the period (1972–2015), 50 articles (11%) combined simulation methods with other techniques. Most of them (70%) mix simulation and optimization, while the remaining combines simulation with a probabilistic model. Discrete event simulation is the preferred simulation approach in 36 papers, while 14 studies used agent-based simulation. WSC mixed-methods publications have tripled in the past six years. Interestingly, the spread of the modeling using different application areas proves the growth in attention to this method.
Mixed-methods articles in relation to decision levels.