01:30 PM - 02:00 PM
Optimality of the Closest-Idle Policy in Advanced Ambulance Dispatching
We address the problem of ambulance dispatching, in which we must decide which ambulance to send to an incident in real time. In practice, it is commonly believed that the ‘closest idle ambulance’ rule is the best choice and it is used throughout most literature. In this paper, we present alternatives to the classical closest idle ambulance rule. We show that significant improvements can be obtained by these alternative policies. The first alternative is based on a Markov decision problem (MDP), thereby constructing the first known MDP model for ambulance dispatching. Moreover, in the broader field of Dynamic Ambulance Management, this is the first MDP that models more than just the number of idle vehicles, while remaining computationally tractable for reasonably-sized ambulance fleets. Second, we propose a heuristic for ambulance dispatching that can handle regions with large numbers of ambulances. For both alternatives, we focus on two performance metrics, namely, the fraction of late arrivals and the average response time. We evaluate our policies by simulating a large emergency medical services region in the Netherlands. For this region, we show that our heuristic reduces the fraction of late arrivals by 18% compared to the ‘closest idle’ benchmark policy. This sheds new light on the popular belief that deviating from the closest idle dispatch policy cannot greatly improve the objective.
02:00 PM - 02:30 PM
Incorporating Coverage for Emergency Calls in Scheduling Patient Transportations
Many ambulance providers operate both advanced life support (ALS) and basic life support (BLS) ambulances. Typically, emergency calls can only be executed by ALS vehicles, whereas non-urgent patient transportations can either be served by an ALS or a BLS ambulance. BLS vehicle capacity does normally not suffice for all transportation requests. The remaining transportations are performed by ALS ambulances, which reduces coverage for emergency calls. Most models in literature ignore this connection between patient transportations and emergency calls. We present a model to determine routes for BLS vehicles, so as to maximize the remaining coverage by ALS ambulances. Since most transportation requests arrive during the day of execution, our model will be online to handle these incoming requests. We show that it is not necessarily optimal to execute the maximum number of patient transportations with a BLS ambulance. Executing fewer patient transportations with BLS ambulances can result in a higher coverage by carefully selecting the transportation requests that are executed by an ALS ambulance.
02:30 PM - 03:00 PM
A matheuristic decomposition approach to solve the dynamic ambulance relocation and pre-assignment problem
Emergency medical services (EMS) generally deal with two real-time decisions, i.e. ambulance dispatching and relocation. Dispatching consists in selecting which ambulance to send to each emergency call, while relocation consists in relocating available ambulances throughout the day in response to changes in the state of the system. Up to now, dispatching and relocation decisions have been generally considered separately. However, they are closely related and joint strategies that consider them simultaneously could help to maintain an adequate service level together with lower relocation efforts.
In this study, we address the Dynamic Ambulance Relocation and Pre-assignment Problem (DARPP) and we propose an integrated programming model to consider the dispatching and relocation decisions together. This global model determines the location of each available ambulance as well as an ordered list of available ambulances that can be dispatched to each demand zone, aiming at minimizing the expected response time as well as relocation efforts. Moreover, to ensure solving real-life instances, a matheuristic approach has been developed by decomposing the problem and exploiting the division of the territory into sub-regions. This solution approach consists of three main steps: (1) ambulances are allocated to sub-regions; (2) the DARPP is solved for each sub-region; (3) the pre-assignment lists are updated based on ambulance locations obtained from (2). Each step consists of a programming model extracted from our global model.
Results obtained from a first set of instances show the benefits of simultaneously considering pre-assignment and relocation decisions, and the impact of pre-assignment decisions on relocation ones. Moreover, on a set of larger size instances representative of a real application context, the proposed decomposition approach is shown to be an effective instrument to provide good quality solutions.