I have read the positions stating that calls for emergency services are completely random (justifying the reason they are often called “accidents”) and therefore not able to be predicted. But both academic literature and practical experience show that demand prediction can be an effective tool in helping to balance scarce resources (ambulances and their trained crews) with public demand (requests for emergency responses even without taking into account the abuses to the system as discussed in a previous posting on the problem of “frequent flyers”) while still improving response times and controlling costs.
For anyone who thinks all of this sounds too good to be true, there are examples of where expensive technology is not having the desired affect. One such location is Lee County EMS in Florida where not only have response times not been improved, but ambulances are burning more fuel than ever and the critics include the very paramedics it is supposed to help. While predicting where the next 911 call will come from may be similiar to “picking the winning card at a casino” as the Florida investigative news reporter suggests, that isn’t really the objective. We don’t need to know which phone will make the next call, it is enough just knowing the probability of a call coming from any given location within the service area. This may be a subtle distinction, but one that makes a huge difference at MedStar in Fort Worth or Life EMS in Grand Rapids where response times were dramatically improved by taking the next step beyond simple demand prediction and placing ambulances at positions where they can be the most effective.
Academic studies show that demand pattern analysis can be used without hourly, daily, or seasonal calibration to achieve potentially acceptable tolerances of demand prediction, but when adjusted with these appropriate corrections, software applications like MARVLIS (the Mobile Area Routing and Vehicle Location Information System) can effectively predict demand in practical situations. According to Tony Bradshaw of BCS, the makers of MARVLIS, it routinely calculates where about 80% of demand will occur and when paired with realistic drive-time response zones it demonstrates valuable support for a dynamic System Status Management plan to pre-position, or “post” ambulances closer to their next call saving valuable time and increasingly expensive fuel costs.
What matters most, though, is what agencies experience in the field. At SunStar they say ” the most significant result was improving our emergency response time from 90.2% to now over 93% in lieu of an increase in patient call volumes. This equates to ambulances arriving on scene more than 1 minute quicker. We additionally saw a savings of $400,000 in penalties by exceeding our contractual goal of 92% and performing above 93% compliance.” Similarly, Steven Cotter, Director of Sedgewick EMS added that “the technology has opened our eyes to be able to understand how we are performing, where we are deficient in our performance and how we can make changes quickly and adapt to a changing environment.” And beyond simple response times, “it’s what technology should do,” says Joe Penner, Executive Director at the Mecklenburg EMS Agency, ” take the complex and present useful, straightforward information. It has helped us improve response times, resource utilization AND simultaneously reduce unnecessary post moves — your patients and employees will appreciate it!”
My conclusion is that proper demand prediction paired with realistic response creates significant opportunity to improve performance and cut costs even in growing communities. When used properly, the future looks bright for High Performance EMS!
High Performance EMS 
I have heard a comment that the word “prediction” is probably not the best term to describe the process of planning for upcoming demand. There is nothing “supernatural” about it, but rather it is a mathematical procedure of forecasting demand based on past history (and the fact that humans are creatures of habit within demographic collections.) From now on, I will use the more precise term of “forecasting” instead of prediction.
It is still true, however, that different forecasting models will provide different results. Research has shown that the MARVLIS demand modeling is clearly among the best in the industry resulting in about 80% of realized demand being properly forecast in about 5% of the service area based on their current customer profiles.