Heterogeneous Facility Location under Coordination and Collaboration: The Case of Ambulance-Bystander-Drone Coordination

Abstract

Drones have demonstrated potential to improve response to time-sensitive emergencies, such as out-of-hospital cardiac arrest, which kills over 300,000 individuals in the United States each year. Drones can facilitate early medical intervention by delivering an automated external defibrillator (AED) before ambulance arrival, but they require a willing bystander to apply the AED. Prior research on designing static or drone-enabled AED networks have assumed that a bystander is always available at the scene, even though historical bystander participation is extremely low. As a result, these models may overestimate system performance and place drones in areas with low utilization. In this study, we introduce the concept of capturing bystander availability in joint location-queuing problems and propose a modeling framework to simultaneously optimize a drone-bystander-ambulance network for cardiac arrest response. Service to each incident can be achieved by either an ambulance alone; or by the drone-delivered AED together with a bystander, complemented by a slower-arriving ambulance. We formulate this problem as a mixed integer linear program and we use the solution structure to develop an accurate and highly scalable heuristic algorithm. We demonstrate our approach on a case study that includes three cities around Toronto, Canada. On real data from the three cities, we find that even a small number of drones (e.g., five) improve response times by 15-20\% and the fraction of calls served within 6 minutes by 27-61%. Simultaneously optimizing an drone-bystander-ambulance network pools ambulances in regions with low bystander availability, while using drones to cover the regions where ambulances are pooled from. Operationally, the drone-bystander-ambulance network is more effective in regions with limited ambulance resources or low ambulance availability. Moreover, adding a small number of drones may be more effective than adding many ambulances, especially in rural regions with lower incidence rates spread over large areas.

Note:
Funding Information: MISSING

Conflict of Interests: MISSING