Optimizing Emergency Medical Services Systems: Integrating Hospital Congestion to Minimize Time to Treatment

Abstract

The efficient delivery of ambulance services is crucial to public health outcomes and all the more urgent in the aftermath of the COVID-19 pandemic's unprecedented strain on healthcare systems worldwide. In this thesis, we formulate two stochastic optimal control problems. The first aims to minimize the expected time to delivery to the hospital. The second also considers congestion at the hospital and aims to minimize the expected sum of ambulance waiting time, ambulance travel time, and hospital waiting time. Running into the curses of the dimensionality, we use an approximate dynamic programming approach to arrive at approximate solutions to our models. We train our models using simulated data from the Fez-Meknes region of Morocco and demonstrate that they can achieve significant performance improvements over myopic heuristics, illustrating the potential of using optimization tools to improve public health outcomes. Furthermore, our results indicate that EMS systems face tradeoffs between the various steps of the patient service process, indicating that optimizing for the efficiency of the entire process simultaneously can improve model performance relative to optimizing for each step individually.