Conference Presentations

EMS Planning Materials

Shane Henderson

Deploying Ambulances: Some Insights and Computational Tools

What fraction of time should your ambulances be busy? If this fraction is too high then response times go up, and if it is too small then you are not effectively using the assets at your disposal. The answer turns out to depend on how large a fleet of ambulances you have. Larger fleets can be run at higher utilizations than small fleets while still keeping response times small. I'll explain why.

How should a given fleet be distributed between remote locations, or between bases? I'll look at a (very) special case of this question, explaining how to distribute ambulances between two locations. It turns out that distributing ambulances in proportion to demand is not always the best answer! It can be better to "boost" the allocation to the lighter-loaded location.

Finally, I'll discuss a new method for system-status management, for example, moving ambulances in real-time to ensure good response-time performance.

I'll explain roughly how our method works and give lots of computational results for (slightly) simplified models of ambulance operations at two cities, one of which is Edmonton.

Armann Ingolfsson

EMS Performance Targets and Travel Times

Most EMS operators use a response time performance target of the form "respond to X% of calls in Y minutes or less." Numbers and definitions vary between operators and regions, but "90% in 8:59 minutes" is a common target in cities in North America. An alternative performance target is to "maximize the number of cardiac arrest survivors." I will discuss possible implications of such a target for how best to choose station locations and deploy ambulances.

Second, I will discuss a relatively simple approach for estimating ambulance travel times, taking into account that average speeds are typically lower for shorter trips, and recognizing that travel times are highly variable.

Jeff Meyer

Siren Live: Software for Real-time optimized Ambulance Redeployment

Emergency service performance is typically measured against the time an emergency response vehicle takes to reach the scene of an incident. Judicious placement of available response vehicles to cover areas with higher expected call arrival rates means the emergency service is more likely to respond to those calls within the requisite response time, hence improving the response time performance of the system. A new optimisation-based decision support tool, Siren Live, makes improved deployment recommendations in real-time, maintaining system performance. It incorporates rules to address staffing issues and other problems that are typically observed in previous methods based on manual heuristic approaches. The Siren Live user interface and optimization model is presented, and experience with implementation is discussed.

Andrew Mason

BARTSim, Siren and Beyond

Discussion of the Better Ambulance Rostering Technology project that incorporated travel models to predict travel times under time-varying congestion. BARTSim later became a more sophisticated system called Siren that uses AVL data and Map Matching Dynamic Programming.

Dan Haight

An EMS Planning Medley - Data Modeling to Improve Operations and Direct Strategy

In this presentation, we will discuss the broad set of structural, strategic, and operational issues that face EMS planners. We will then look at specific case studies of how the different issues can be addressed. Finally, we will identify some rules of thumb with an eye toward developing a cost-benefit framework for EMS.