Our study is the first to compare widely used modeling assumptions with actual EMS trip records within a Canadian setting. A comparison of the modeling assumptions with true EMS data showed three key differences. The first was that the on-scene time in our study area was higher than that reported by the US meta-analysis. The second was that the empirical multiplier, used to derive the response time from transport time, was higher than previously reported in the US, for areas outside the immediate urban boundaries of the City of Calgary. The third was that the actual EMS pre-hospital times across our study area were significantly higher than the estimated times modeled using GIS and the original travel time assumptions. Researchers undertaking studies that evaluate the access of populations to urgent services by ground ambulance should consider that the previous assumptions may not be indicative of typical time intervals in their jurisdiction. Thus, in order to represent the pre-hospital time in our study area more accurately, we created a revised model specific to our jurisdiction that successfully reduced the overall difference between actual and modeled pre-hospital times.
The higher on-scene times we found in our study area could be applied to models across Canada. Another recent study found that the on-scene interval in Canada was similar to that noted in our study (median: 20.2 minutes; Inter Quartile Range (IQR): 14.9–27.0) showing that this interval may be consistently higher than measures previously reported in the US. The debate on whether to ‘scoop and run’ or ‘stay and play’ once EMS reaches the scene has been on-going for almost two decades with mixed recommendations[18–20]. The choice of protocol used in an area will ultimately affect the scene times for a particular jurisdiction. The larger empirical constant used to derive rural response times in our study (0.89 vs. 0.4 in previous studies[5, 6]) is a reflection that ambulances in areas outside the city require a greater amount of time to reach the patient scene from their originating locations. A recent study has shown that areas with low population density are at risk of delayed ambulance response times. Overall, it appears that the underestimation of on-scene and transport intervals inside the city as well as an underestimation of the response intervals for areas outside the city results in the underestimation of total pre-hospital time. This underestimation of pre-hospital time is especially noticeable in areas surrounding major roadways as seen in Figures4 and5.
Although there have been no studies that have validated this commonly used method for evaluating pre-hospital time by ground ambulance, there have been studies that validate the GIS modeled transport interval. One study, conducted in a large Canadian city, calculated ambulance driving times using GIS for critically injured patients and found that the true transport intervals were more variable than the modeled transport interval. For a single origin-destination pair they found actual transport times to be between 8 and 27 minutes, while the GIS modeled time remained constant at 13 minutes. This was similar to our study findings where the range of actual travel times was greater than the single GIS transport time interval calculated for a given scene to hospital trip. This shows the importance of recognizing that all pre-hospital time estimations anchor on measures of central tendency. The value of the modeled time is that it provides a statement of expected time between scene and hospital locations, but in the real world there will always be a dispersion of values around the average. This type of estimate is useful because it provides decision makers with information regarding access to specific services. Collecting, cleaning, and analyzing historical data over large areas (e.g. across regions or countries) would be exceptionally time consuming when multiple EMS service providers are involved. Using empirical constants with GIS to model pre-hospital times can greatly facilitate a regional or national assessment of access to health care services by ground ambulance.
Evidence suggests that ambulance pre-hospital times around the world vary. The Canadian city of Montreal, Quebec reported a median response time of 8 minutes, a median on-scene interval of 16 minutes and a median transport interval of 9 minutes. Monterrey, Mexico reported a median response time of 4 minutes, a median on-scene interval of 10.1 minutes and a median transport interval of 5 minutes. While the city of Urmia in Iran reported on-scene intervals that were notably shorter than those reported in the US and Canada (median 5 minutes (IQR:4–7) within the city, and 7 minutes (IQR: 5–11.3) outside of the city). These pre-hospital intervals reported from different regions of the world show that a generalized model across different countries may not be appropriate. Even across the US, the pre-hospital time intervals have been shown to vary. Because different jurisdictions have different geographic barriers and different EMS protocols, the travel time modeling assumptions should be adapted to the study areas under consideration.
There are limitations to creating a generalized pre-hospital travel time model. Real time traffic conditions and weather can add significantly to the response and transport interval portions of the pre-hospital time. In our study we used an entire year of data, which included EMS trips conducted at different times of day and during different seasonal conditions, to create generalized models for pre-hospital times, which incorporated these variations. When measures of access need to be considered over more specific conditions, data could be stratified by times of day, days of the week or by months of the year. Our goal in this study was to create a generalized model comparable to those used to measure national levels of population access to urgent care[8, 10, 12], which allows for an estimation of patient pre-hospital time in the absence of individual EMS historical data.
Based on previous methods we derived a relationship between the response and transport intervals to account for the time from ambulance location to scene across the study area. The linear multipliers used in previous studies and applied here in our revised model may not be sufficiently sophisticated to capture the relationship between the response and transport intervals. The exploration of new mathematical relationships between time intervals is a potential future direction of this research. It is also important to note that there are factors that are unaccounted for that can affect this relationship. Studies have been conducted showing that higher call volumes and higher intervals of vehicles unavailable for response can increase the response time. Improvements to ambulance deployment and changes in the demand volume could reduce response intervals in an area, which would require further revision to the models proposed in this study. When using generalized models across a country it is important to recognize that EMS systems in different jurisdictions have unique protocols[25–27]. The geography of an area could also affect response times. For example, areas with similar urban structures (e.g. urban sprawl) may be affected by delayed response times.
Finally, the reduction in the overall difference between actual and modeled pre-hospital times with our revised model is not surprising considering that data from our study area was used to create this updated model. To truly understand the effectiveness of this model we need to apply it to different jurisdictions and evaluate it in relation to actual EMS trips undertaken in these areas. In spite of these limitations we believe that the estimation of pre-hospital time using GIS is valuable for studies focused on access to urgent care, especially in the absence of actual EMS trip records. There are several messages for other jurisdictions to take away from this study. Different areas have different EMS protocols and pre-hospital benchmark goals in place that affect their activation, response, on-scene and transport intervals. It is expected that within a single country, different jurisdictions could have significantly different median pre-hospital interval times because of their unique EMS protocols. The question of which model or assumptions to use is a question of both Geography and policy.