Inhomogeneous Poisson Process for Ambulance Dispatch

This paper introduces a method to select the most appropriate ambulance to respond to an emergency. For vital emergencies, the proposed method dispatches the closest available ambulance. For non-vital emergencies, our approach is based on risk assessment. Specifically, we prioritize the vehicle that minimizes this risk among all ambulances whose dispatch would increase the response time by less than a given threshold. The notion of risk here aims to measure the probability that a vital emergency occurs while there is no available ambulance at the closest station. To evaluate the proposed approach, historical data spanning six years provided by the Centre Hospitalier Universitaire Vaudois located in Switzerland have been used. Data from 2015 to 2020 have been utilized to build our method and the data from 2021 for the evaluation. Experiment design choices, such as leveraging an API for travel time estimation, the use of real-world data, and the incorporation of resource relocation, make our six experiments representative of real-world situations. The consistency of the results across all experiments suggests that our method could be successfully used to reduce the number of ambulances unavailability at the closest station for vital emergencies. This reduction consequently leads to a decrease in the average response time for vital emergencies or a reduction in the number of ambulance relocations during the days. These observations collectively suggest that our method has the potential to enhance the quality of emergency medical services and, consequently, save lives.