Infrastructure Planning Models for EMS

Principal Investigator: Armann Ingolfsson, Ph.D., Academic Director, CEO

Project Manager: Dan Haight, B.Comm. (

Project Sponsor: Infrastructure Canada, Peer Reviewed Research Programs

Project Summary

The project aimed to develop and apply mathematical models to analyze Emergency Medical Services (EMS) operations, to shed light on and influence policies regarding the allocation of EMS funding between regions, strategic station location, scheduling, and system status management.

Le projet a été commissionné pour créer et appliquer des modèles mathématiques pour analyser les opérations d'Aide Médicale Urgente (AMU). Ces modèles ont aidé en illuminer et en influencer des politiques concernant l'attribution du placement de AMU entre les régions, localisation stratégiques des stations, l'établissement du programme et la gestion d'état de fonctionnement du système.

Areas of Focus

The creation of mathematical models to predict the impact of greater integration of the EMS systems in the Capital Health region under various scenarios.

Strategic Station Location
Long-term planning of where and when to build (or close) ambulance stations in response to increases (or decreases) in population and demand.

System Status Management
Modeling of SSM policies and their impact on response time performance as well as crew fatigue.

Executive Summary


In March 2004, the Alberta Government decided to transfer EMS governance and funding from municipalities to the regional health authorities. The transfer will take effect in April 2009. Regionalization makes it important to consider how EMS funding is or should be allocated between regions and how to operate a regional EMS system. A strict population-based allocation ignores many important factors that impact the costs and operations of such services and would result in large variation in EMS access between regions.

Working with the Capital Health Region (which encompasses the City of Edmonton and surrounding urban and rural areas) we collected, cleaned, and standardized data from five EMS operators in the region (Edmonton, Leduc, Parkland, St. Albert, and Strathcona County). We worked with the operators to choose performance benchmarks for urban and rural areas and to operationally define boundaries between urban and rural areas. Then we measured performance for each operator, separately for urban and rural areas, and prepared various graphs and maps to improve understanding of system performance and how it could be improved. In addition to measuring and diagnosing current performance, we used mathematical models to predict the impact of greater integration of the EMS systems in the region under various scenarios.

We used statistical and mathematical models developed by us and other researchers to predict future performance under various regionalization scenarios, including (1) ignoring all municipal boundaries for EMS dispatching purposes, (2) reduced hospital drop-off times because of better coordination between EMS operation and the health region, and (3) adding EMS vehicles in selected areas of the region. The predictions specified how the overall system performance might improve, as well as which geographical regions would benefit the most.

As part of analyzing the data from the five EMS operators, we continued to develop an "overgoal classification" method, related to fault-tree analysis in risk analysis and fishbone diagrams from total quality management, to help diagnose EMS system performance and identify areas for improvement. The overgoal classifier attempts to identify a root cause for every EMS call whose response time exceeds a specified target (commonly 8-10 minutes in urban areas and 20-30 minutes in rural areas).

Station Location

EMS operators must constantly evaluate the need to add or remove EMS stations, in response to or anticipation of increases, decreases, or geographical shifts in demand. As part of the project, we advanced mathematical models for optimal station location. Our aim is to develop tools to optimally locate stations in a municipality over a 10 - 20 year time horizon. With regionalization, this work has taken on added importance, because EMS operators in some parts of the province now have the freedom to design their systems "from scratch" without being bound by investment in existing stations.

Our base model for this work chooses facility locations to maximize expected coverage (fraction of calls reached within a response time standard), assuming unlimited ambulance availability at every station. The assumption of unlimited ambulance availability is of course unrealistic, but it leads to a more tractable model and it decouples strategic decisions about where to locate stations from shorter-term decisions about how to allocate ambulances to stations.

We have been successful in solving the base model for the Calgary EMS system at a high level of aggregation (dividing the city into 1453 "dissemination areas") but solving the multi-year problem is likely to require aggregation of demand. We have successfully solved test problems where demand is aggregated from 1453 to 500 demand nodes and a 15-year time horizon is aggregated into 3 representative years (Year 5, Year 10, and Year 15). In future research, we plan to investigate the impact of such aggregation on the accuracy of the model results and methods to reduce the loss in accuracy.

System Status Management

Also known as "roving deployment," "real-time repositioning," "redeployment," or "flexing," System Status Management (SSM) involves dispatching decisions about how to reposition ambulances to maintain coverage, especially when the number of available ambulances in the system is low. Such repositioning increases crew workload and therefore involves a trade-off between crew fatigue and response time performance.

SSM has a large impact on response time performance. A study we completed for Edmonton EMS in 2005 indicated that "turning SSM off" would reduce the fraction of calls reached in 9 minutes from 87% to 81% and that it would require 8 new ambulances to match current system performance in the absence of SSM.

We have made considerable progress in developing a tractable Markov chain model to evaluate SSM policies. We are continuing with both theoretical and computational analysis of this model, including validation against the discrete event simulation model and field data from Edmonton and Calgary.

Academic Papers

Peer-reviewed articles on topics related to the project that were submitted, accepted, or published during the project.

Budge, S., A. Ingolfsson, E. Erkut. 2008. Optimal ambulance location with random delays and travel times. Health Care Management Science 11 262-274.

Budge, S., A. Ingolfsson, E. Erkut. 2009. Approximating vehicle dispatch probabilities for emergency service systems with location-specific service times and multiple units per location. Operations Research 57 251-255

Budge, S., A. Ingolfsson, D. Zerom. 2008. Empirical analysis of ambulance travel times: the case of Calgary Emergency Medical Services [submitted, last revised November 2008]

Channouf, N., P. L'Ecuyer, A. Ingolfsson, A. N. Avramidis. 2007. The application of forecasting techniques to modeling emergency medical system calls in Calgary, Alberta. Health Care Management Science 10 25-45.

Erdo?an, G., E. Erkut, A. Ingolfsson, G. Laporte. 2008. Scheduling ambulance crews for maximum coverage. Journal of the Operational Research Society, forthcoming.

Erkut, E, A. Ingolfsson, S. Budge. 2008. Maximum availability/reliability models for selecting ambulance station and vehicle locations: a critique [submitted, last revised May 2008]

Erkut, E., A. Ingolfsson, G. Erdo?an. 2008. Ambulance deployment for maximum survival. Naval Research Logistics 55 42-58.

Erkut, E., A. Ingolfsson, T. Sim, G. Erdo?an. 2009. Computational comparison of five maximal covering models for locating ambulances. Geographical Analysis, 41 43-65

Media and Press

Calgary Herald Newspaper
Western Wheel Community Newspaper
City of Calgary Press Release

Project Impact

In addition to the outcomes discussed for the individual research areas above, we organized a planning conference for EMS practitioners and researchers to be held in Edmonton from August 7-8, 2008. The conference featured internationally recognized speakers and EMS practitioners from all parts of Canada were invited to participate.

The EMS Planning Conference was aimed at EMS managers and analysts with the goal of sharing knowledge about effective planning and management of EMS systems. The first day of the conference provided training for EMS analysts. The second day featured presentations by EMS researchers, as well as opportunities to influence the direction of future research.

Both practitioners and academics attended our EMS conference. The keynote speaker, Dr. Andrew Mason from Auckland, New Zealand, talked about his recent work in System Status Management.

Other presenters included Dr. Armann Ingolfsson (Travel Time Estimation), Daniel Haight (Practical Applications of EMS Modeling), Dr. Shane Henderson from Cornell University (Resource Allocation), and Jeff Meyers from Optima Corporation (Using Simulation Tools to Improve Operations). We facilitated two-panel discussions (Data Accuracy in EMS; Regional Collaboration) and offered a full day data analysis training seminar to EMS practitioners.

Conference Presentations

This project has been made possible through a financial contribution from Infrastructure Canada.