Evaluating Emergency Medical Service Locations by Response Time Distribution and (Conditional) Value at Risk
Abstract
When evaluating the ambulance sites of emergency medical service (EMS) in a stochastic context, location planning often considers the expected coverage value only. However, comparing the total distribution functions of response time of all emergency calls might provide additional information. In the context of stochastic programming, the performance criteria value at risk (VaR), conditional value at risk (CVaR), and stochastic dominance are used for the evaluation of uncertain situations. Thus far, the above mentioned performance criteria rarely have been applied in the context of EMS. We therefore propose to evaluate decisions with regard to strategic EMS location planning using the total distribution functions of response time of all emergency calls and applying VaR, CVaR, and stochastic dominance. We use different coverage and fairness objectives of a current stochastic location planning model in order to examine the performance criteria’s applicability. Based on a local EMS provider’s anonymized real-world data we perform a case study, and conduct a discrete event simulation to obtain the individual distribution functions. We find that the VaR emphasizes the gap between the coverage objective and the achieved actual coverage. Thus, it is advisable to use the VaR to evaluate the EMS’ long-term development and goal achievement. Although we show that the selected performance criteria represent coverage objectives better than fairness, we demonstrate that it is still important to additionally consider fairness in order to improve the system quality for the least covered areas.