Research on the impact of the location of EMS stations on EMS response time began almost 40 years ago. As the scope of the interest in EMS expanded, researchers were impelled to examine other spheres excluding location, which affect the response time. These included predicting the number of resources needed to obtain a specific average response time. For instance, Baloyi et al. (2017) predicted the number of emergency units required but, failed to predetermine their location. Moreover, over the past years, there has been a growing body of research that has focussed more on the geographical accessibility for various EMS locations. This inclination was instigated by the need to evaluate ambulance service levels. Ambulance response time to specified locations has been identified to be among the primary factors affecting the levels of ambulance service (Geertman et al. 2003). This was mirrored in Baloyi et al. (2017) in which the analysis approach was used in examining ambulance response. In the US, previous studies on ambulance response have been used as a baseline for guiding recent studies in modeling national access to services by ground ambulance (Branas, C 2005; Carr et al. 2006). The use of the geographic information system (GIS) is essential in modeling travel time along with a road network. However, some studies failed to use this technique (Alrazeeni et al. 2016; Hsia et al. 2017). Others such as Maghfiroh et al. (2018) and Mansour (2016) have done otherwise. A textbook example of the use of GIS is in Patel et al. (2010) in which pre-hospital time assumptions were complemented with GIS to assess Canada’s population access to Percutaneous Coronary Intervention facilities.
Models used in emergency routing
Many authors have examined the impact of GIS systems on EMS response time, through which several developed emergency routing models. Arentze and Timmerman (1996) developed a model that was grounded on establishing the shortest course from accident points to the locations from which ambulances are dispatched. The model was also based on finding the shortest route from crash locations to healthcare facilities (Arentze &Timmerman 1996). the population of zone improvement plan (ZIP) regions was used to drive the searching of service coverage, therefore, enabling the researchers to determine the shortest path through minimizing the overall distance from EMS units to crash points, and then to hospitals.
Furthermore, this was reflected in Church, RLJGis (1999) in which the location set covering model was developed to minimize the number of emergency bases attending to various emergency incidents in different locations. The location set covering approach aimed to ensure the provision of maximum service coverage with a rigid number of EMS units. On top of using accessibility as a measurement in representing the number of emergency service units needed to provide service, the number of EMS locations identified can facilitate the determination of the population that each unit can attend to within the response time standard (ReVelle, CS & Eiselt 2005).
Radke and Mu developed the spatial decomposition model Radke and Mu (2000). This model was created to assist in the prediction of spatial accessibility to social services. It functions to ensure that the public, including the under-served areas, have equal access to emergency services. The model achieves this by determining the availability and accessibility of these services using the point-to-point and point-to-polygon approaches to the service coverage. The point-to-point approach determines the shortest distance from the crash point to the nearest emergency service units and hospitals, however, the point-to-polygon determines the travel time and uses it to develop a service area polygon. Nevertheless, one of the main challenges affecting emergency routing is the coupling of the data obtained from the two before mentioned approaches with EMS data, for instance, road traffic accidents. Radke and Mu (2000) utilized the polygon-to-polygon technique to locate a fixed supplier from which buffer polygons were created at a 60-mile radius. Moreover, they downloaded the household densities from the census tract and integrated this information into their polygons. The household density present in the polygons was in the form of data that was disaggregated utilizing the buffer of each supplier’s coverage area. Therefore, through this, they were able to calculate the accessibility in overlapping regions that were served by several contractors within a response-time threshold. In summation, the spatial decomposition model outlines the level of emergency service, that is whether under-served or over-served, that a spatial pattern of a population receives about other regions.
Last but not least is the heuristic substitute model. This model functions to relocate the suppliers, in other words, the capacity-layer location of ambulance stations, and create buffers. Therefore, this will ensure that all processes in every region have robust service probability-coverage maps. Luo, Wei et al. (2003) made use of GIS in their study on the spatial accessibility of service areas. In the study, service areas were defined by the travel time about the availability of physicians and the capacity of trauma centers.
Importance of response times
Over time, the value of response times as the primary factor affecting emergency response time has been appreciated such that further research has been conducted to examine the effect of rapid response in medical emergency services (Branas, CC, MacKenzie &; ReVelle 2000; Burger et al. 2018). For instance, Al-Ghamdi and Prevention conducted a study to evaluate ambulance rescue time in Riyadh (Al-Ghamdi and Prevention 2002). The objective was to be achieved through analyzing the time for highway crashes and comparing the rescue time in Riyadh to that of other countries. The ambulance rescue time included the response time, time at the crash point, and time is taken to reach the hospital. The results illustrated that the average rescue time was 35.85 min, in which the mean response time was 10.23 minutes, and the ambulance time at the crash point in conjunction with the travel time to the hospital accounted for the difference. Moreover, Al-Ghamdi and Prevention reported that the ambulance speed to the crash point had an average mean of 55.05km/h with a standard deviation of 27.42km/h.
In another study carried out in Tehran, Panahi and Delavar (2008) aimed to determine the shortest dynamic path in traffic networks, in which the arc travel time fluctuated with time. It proposed the use of a dynamic routing system that revolves around the incorporation of GIS in real-time traffic situations. GIS was suggested mainly because it improved the visualization of the urban network map and the evaluation of emergency routing. Moreover, GIS functions as a powerful tool in finding the shortest path in real-time. However, the study did not consider hospitals.
Blackwell et al. (2009) examined the ambulance response time, providing clinical care, and patient health outcome for high acuity 9-1-1 calls to establish if the present response time requirements established for the urban metropolitan jurisdiction are safe. The research was grounded on the relationship between the time duration at which a 9-1-1 call was received at the dispatch center, the arrival of the ambulance at the scene, and the patient outcome. Moreover, the study focused on the difference between two groups, that is, the control group and study patients or cases. Mean response time of 10 minutes and 59 seconds obtained from the cases was used to benchmark the control group. Therefore, the control group was defined by a response time of ≤10:59 minutes. The results were analyzed using statistical techniques (Blackwell et al. 2009).
Preferably, emergency response times should be minimal for every 9-1-1 call. However, specific calls are more time-sensitive than others as each second counts, for example, emergency cardiac incidents. Ong et al. (2010) employed a prospective observational study to help establish if a GIS-centered-deployment strategy could minimize the reaction time for out-of-hospital cardiac arrests (OOHCA) in an EMS system located in an urban environment. The study was based in Singapore and involved an analysis of geographical locations of all cases of OOHCA. The GIS was used to spot-map these locations. After that, satellite ambulances were progressively deployed, and this resulted in the increase of ambulance stations from 17 to 32 localities. The study also took into consideration the discrepancy in the deployment of ambulances relative to demand and hour of the day. It is essential to note that during the study period, the proportion of ambulances and crews remained steady. The results showed that the response times at 8 minutes increased from 22.3% to 47.3%, while that of 11 minutes rose from 57.6% to 77.5%. However, the study did not use the international standard of time.
Moreover, to account for the differences in rural and urban environments, Gonzalez et al. (2011) researched the rural county of Alabama to assess if the repositioning of emergency service stations could lead to the improvement of EMS response time to motor vehicle crashes (MCVs) without negatively influencing the response time to non-MCY-related cases. The study used GIS software to assess the locations of various MCVs over nine months. It was established that there was only a single ambulance agency serving the whole county. The researchers geographically divided the county into two regions relative to concentrated areas of MCVs and a new emergency station was assigned to every area. However, the proportion of ambulances in-service remained constant. Redeployment data, that is, regarding the emergency response time to the scene, miles covered by ambulances to reach the crash point, fatalities, and type of accident (MCV or non-MCV) was collected. This data was then compared to non-redeployment data, in other words, real-time historical data, collected from 9 months in a similar time. The results indicated that during the redeployment period, a total of 597 9-1-1 calls were received, of which 17.8% comprised of MCVs. On the other hand, in the non-redeployment data, MCV calls constituted 8.1% of the 764 EMS calls documented. Furthermore, during redeployment, the mean distance that ambulances traveled to a crash point was 8.6 miles, unlike 10.7 miles before redeployment. Lastly, the mean duration to a crash point was 8.0 minutes all through redeployment as compared to 9.5 minutes before redeployment (Gonzalez et al. 2011).
Last but not least, Jezek et al. (2011) proposed a technique of estimating ambulance response times relative to a classified road network. This method took into account different attributes of road networks, such as road category, type, class, and surface. A comparison of the calculated and measured times illustrates that the estimated response times would probably have no favorable outcomes in real-life scenarios. The authors complemented the obtained values by making use of the GIS to make comparative judgments regarding estimated and historical time. The judgments revolved around the establishment of the shortest and fast path, air distance methods on the ground ambulance, and other factors, such as weather and rush hour that might affect response time (Jezek et al. 2011).
Some studies have focussed on the “Gold hour”, which is regarded as the first hour when trauma or a condition occurs. For instance, the study in (Branas, CC, MacKenzie & ReVelle 2000) aimed to estimate the proportion of US residents that have access to emergency centers within a 45-60 minutes radius in all the 50 states. It evaluated both the availability and accessibility of air and ground ambulance while ignoring the effects of response time (Branas, CC, MacKenzie & ReVelle 2000). Another research utilized the 45 minutes as a benchmark on which the goal was to reach 90% of the population. (Zakariassen, Uleberg & Røislien 2015) carried out a study in Norway that documented accurate flying times for air ambulances to arrive at the scene and determined the rates of acute primary missions.
Nevertheless, some studies infringe more on the emergency response time as they compare the measured time to that of the international standard, which is regarded to be 8 minutes for a case study. For instance, in Alnemer et al. (2016), confounding factors such as time, gender, age, and location of incidents for 18 stations across Riyadh were evaluated. It was reported that the mean response time was approximately 13 minutes, with an inclination towards prolongation, especially on weekdays during working hours. Moreover, ART (Average Response Time) and survival rate were only affected by the age of the patient, and not the gender of the patient or location of the incident. Therefore, Alnemer and his colleagues substantiated that ART was much longer than the 8 minutes defined by the international standard of time (Alnemer et al. 2016). To add to this, most prolific research papers measure time, consider the rush hour or the international standard of time (Amorim et al. 2019; Baloyi et al. 2017; Lee 2014; Maghfiroh et al. 2018; Patel et al. 2012; Saba, Noor & Malik 2017).
Overall, previous research shows that ambulance agencies do not provide an equitable service level and response time in their respective areas of coverage. However, there is room for improvement, especially in cases where health conditions such as OOHCA are linked with the response time. Furthermore, in such cases, emergency agencies need to embrace the “chain of survival” which suggests that survival can be maximized through early access, timely cardiopulmonary resuscitation (CPR), timely defibrillation, and prompt advanced care. Currently, there is substantial research reporting the value of delivering early defibrillation. Therefore, about the above loopholes detected in previous research, the primary objective of this study is to establish whether a deployment strategy based on GIS can minimize emergency response times in an urban environment. Additionally, this study will be benchmarked on a 4-minute model. No study has used the 4-minute model or compared the 4-minute to the 8-minute model.