Theoretical modeling of spatial accessibility in the management of stroke in the Rhône department (France) and comparison with measured data

Stroke is a leading cause of death and a major cause of irreversible sequelae. Stroke patients must be transported as fast as possible and their management should be as effective as possible. Firefighters transport a great proportion of these patients. Using a Geographical Information System can be a great tool to measure and predict travel times during stroke patient management. This study modeled travel times from the nearest fire station to stroke patient using theorical data sets in order to compare them. The results were then compared with the on the job data measured by the Departmental Metropolitan Fire and Rescue Service (SDMIS) from 2015 to 2016 for suspected stroke. This comparison assessed the feasibility of using a theoretical model to predict travel times based on real-life data. A strong correlation was observed between the different theorical models for measuring accessibility to the nearest fire station, with a Root Mean Square Error (RMSE) about 1.5 min and Mean Absolute Percentage Error (MAPE) around 30%. However, when theorical data sets were compared to measured data, the correlation was lower, the RMSE was 6 min and MAPE from 60% (minimal times) to 25% (median times). This is satisfactory results. Although the coefficients of correlation remain low, due to the high heterogeneity of the measured times. These differences of few minutes represented a very small portion of the stroke patient's care pathway. Theoretical models were highly correlated and the correlation with the measured data was mostly correct. The few minutes difference between theoretical models and measured data could be explained by traffic hazards and organizational vagaries that always tend to disrupt modeling results. Using measured data was found to be very useful to perform theorical model and to develop a robust model of stroke transportation.