Travel time estimation for emergency medical vehicles with applications to location models.

摘要

Significant effort is expended by ambulance providers in planning, equipping, and staffing in order to meet performance goals based upon proportions of on-time responses achieved, which is one of the most important indicators of system performance used in the industry. Since travel time is usually the longest segment of time taken by an ambulance to respond to the scene of an incident, existing travel time models are evaluated and both simple and complex travel time estimation models are developed. On-time responses are addressed through failure distance models using a survival analytic framework. In turn, these survival models are then used within a location model construct to optimize vehicle placement in order to maximize on-time response coverage. Even though the emergency medical services (EMS) industry has received substantial attention by location modeling researchers in, among other places, the operations research literature and geographical analysis literature, few related studies have explicitly addressed travel time, an unusual situation given the importance of distance metrics in the aforementioned modeling disciplines. Results from the research into travel time and response time estimation and failure distance models are applied in location models to examine how they may affect solution patterns and levels of coverage. The data used come from published travel time studies and from EMS providers. The major findings of this research are that simple travel time models are adequate and additional variables do not significantly add to the understanding of variation in travel time. In addition, average travel speeds are not constant between emergency responses for service and the probability of arriving "on-time" decreases as distance increases. This is the first work to model response probabilities as function of distance and demonstrate that such probabilities can be estimated by a theoretical Weibull survivor curve.