Year: 2013
Type: dissertation
Abstract: When an incident occurs and we need acute health care, we call the emergency number.After such a call, an ambulance should be with us within 15 minutes.To this end, there are ambulance stations all over the Netherlands, from where ambulances can depart to the scene of an incident.In this research we determine the optimal locations of these stations, to be able to provide the best coverage with the least number of stations.In 2010 there were 203 ambulance stations in the Netherlands.The research was carried out at the National Institute for Public Health and the Environment, the RIVM, which advises the ministry of Health, Welfare and Sports on for example ambulance care.We consider the Netherlands as a directed graph, where the nodes represent four-digit postal code areas and the arcs represent the road network, and each arc has a value corresponding to the travel time between the tail node and the head node of the arc.We look at different models, which are all based on two models.The first model is the Location Set Covering Model (LSCM), which determines the minimal number of location sites needed to cover all demand points.The second model is the Maximal Coverage Location Problem (MCLP), which determines the maximal coverage that can be achieved with a fixed number of location sites.With the LSCM we determine the minimal number of ambulance stations needed to cover all areas using three different maximal travel times of 8, 12 and 15 minutes.As a result we see that if we consider the Netherlands as one region, we need 302, 165 and 100 location sites respectively.If we consider each RAV as a separate region we need 203 location sites with a maximal travel time of 12 minutes.Using the results from the LSCM, we use the MCLP to determine the percentage of the residents we can still cover with less ambulance stations than the outcome of the LSCM.We do this for the same three different travel times.We see that we can still cover approximately 95% of the residents with 225, 95 and 60 location sites respectively.Next we again use the LSCM and MCLP, but now distinguish between urban and rural areas, assigning a different maximal travel time to the two different areas, namely 8 and 12 minutes and 12 and 15 minutes.We see that we then need 217 and 173 location sites respectively to cover all demand points.Next, we include the travel time from the incident to a hospital and use an MCLP-like model.We look at two sets of hospitals: all hospitals with an Emergency Department (ED) and the Trauma Centers.These sets consist of 98 and 11 hospital locations respectively.We observe that we can get 99,9% of the residents to a hospital with an ED and 83,8% to a Trauma Center within 45 minutes (and reach them within 12 minutes travel time after the call), if we use 165 location sites.Finally we introduce the use of rapid responders, which can only provide the acute care, where an ambulance might come later if the rapid responder is already there.The results indicate that the use of these cheaper rapid responders could be advantageous.v vi SUMMARY In all the models we have looked at a greenfield scenario, where we can build an ambulance station anywhere, and assume there are no existing ambulance stations.Furthermore we consider one non-greenfield scenario for the LSCM and MCLP, where we already fix an ambulance station at each of the nine different locations of the academic hospitals.Even though these results give several indications, some of the results are inconclusive due to the fact that we considered four-digit postal code areas, which represent city districts or small villages.To get a better insight we recommend to work with for example six-character postal code areas, which represent streets.A consequence, however, is that the computing time might increase to a level that makes it difficult to determine provably optimal solutions within reasonable time.Further research to obtain good approximations of optimal solutions is recommended.
Author: D. Looije
Institution: Delft University of Technology
Language: English
Field: Decision Sciences
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W18587680
