A probabilistic approach of the single vehicle accident’s severity

Abstract

Single vehicle accidents (SVAs) arouse the interest not only of researchers, but also of the European Commission and other international road safety bodies, since the third part of all road fatalities from Europe is caused by the single vehicle accidents. The aim of this paper is to assess the severity of such accidents by estimating the probability of the fatalities (Pd1) and major injuries (Pd2) generated by the single vehicle accidents and to identify the factors affecting those probabilities. As for this research, a complex 6 -year- accidents data base has been used and the accident records have been aggregated on a daily basis. A binary multiple logistic regression has been developed for each type of severity (fatality and major injury) using 86 predictors related to the place of accidents, road category and feature and characteristic, the number and the width of the lanes, horizontal road markings, safety components of the road, road surface characteristics and adherence, weather and lighting conditions, vehicles mileage and drivers’ sex. The logistical models have been tested on their statistical significance and their explanatory efficiency was discussed. Descriptive analysis has been conducted for both models in order to discuss the distribution of the probability values. Pd2 model has better explanatory power than Pd1 and its overall percentage of the predictions is 96.10 %. It is also has a very good homogeneity since all its predictors have positive values. An interesting finding is that no other predictors related to weather or lighting conditions do significantly explain the probabilities of a SVA to be of fatality or to generate major injuries. These constraints are to be further researched since the daily level of data aggregation studied in this paper influence the “immediate effect” of random phenomena, like weather or lighting conditions. Pd2 model has good applicability in identifying, prioritizing and treating the black spots. The Romanian road authority could also use it in driver education and injury risk identification in order to mitigate the severity of the accidents.