Estimating an Injury Crash Rate Prediction Model based on severity levels evaluation: the case study of single-vehicle run-off-road crashes on rural context

Abstract

Abstract In general in case of crash situations the quality of collected data is very limited and several information are usually unreliable. Thus it is recognised that a significant effort is required in order to improve the quality of the crash prediction models moreover a crucial role is played by the identification of the factors influencing the crashes occurrence and the levels of severity estimation. In this paper two injury crash rate prediction models related to single-vehicle run-off-road crashes type are calibrated and in particular significant attributes estimated are identified not only with roadway geometric characteristics and surface conditions, but also with gender/number-of-drivers. To this aim a survey of injury crashes on two-lane rural roads collected in the Southern Italy was considered and analysed. Finally before the calibration step, a preliminary analysis of the data was provided through the estimation of the levels of severity by multinomial logit; in fact by this model only segments with highest values of severity are identified and involved in the calibration procedure.