Discrete-Event Simulation

Simulation Modeling

In general, models allow researchers to explicitly explore the elements of a decision/problem and mediate understanding of the real world by rendering it comprehensible. Simulation modeling is any activity where the actual or proposed system is replaced by a functioning representation that approximates the same cause-and-effect relationship [Page 399]of the “real” system. Simulation allows researchers to generate evidence for decision making or to develop understanding of underlying processes in the real world when direct experimentation (due to cost, time, or ethics) is not possible. Experimentation with simulation models is performed through sensitivity analyses, where the parameters of the system are varied, or through what-if experiments, where the number or types of resources of the system are varied.

Decision trees and Markov models have, to date, been the most common types of computer simulation models used in healthcare. These methods are used to create highly structured representations of decision processes and alternative strategies. This is done by constraining the formulation of these models to a limited vocabulary, essentially three building blocks—decision nodes, chance nodes, and outcome nodes. The main advantage of this type of formulation is that the highly structured format is relatively transparent and easy to interpret. The disadvantage is that the highly structured framework restricts the types or problems that can be articulated, often forcing significant compromises on the model and the modeler. With over 100 building blocks, DES has a much broader vocabulary (than tree models), allowing a broader array of problems to be modeled, with fewer compromises. This means that though there is more to learn initially there, is a greater range of problems one can model.

...