Estimating the performance of emergency medical service location models via discrete event simulation

Abstract

We address the problem of evaluating deterministic EMS (Emergency Medical Service) location models via a simulation approach. For deterministic set covering location models, the performance of the model is typically determined by an objective function representing a certain type of coverage. After determining the location of EMS stations by deterministic models, we propose to conduct a simulation analysis to evaluate the performance by estimating the “real” coverage of the population that takes into consideration the unavailability of the busy ambulances. By using optimization tools, we find the location of ambulances for each model by solving the mathematical models and then we simulate each setting for two different policies under the same parameters. The overall performance of the models is firstly tested on Istanbul data and then on randomly generated data with different problem size, different layout and different arrival rates. We demonstrate that the performance estimated by the simulation based assessment can vary with respect to the problem characteristics and that the performance of a seemingly good deterministic model may not be that good as measured by the simulation approach.