Investigating Injury Severity Of Road Traffic Accidents Involved Motor Vehicles Using Random Parameter Model

According to the Global Status Report on Road Safety 2018 issued by the World Health Organization,deaths and injuries caused by traffic accidents remain to be a global serious problem.Besides,traffic safety situation is still severe in China.In 2017,the number of accidents caused by motor vehicles takes up 89.8% of all the accidents,and thesecrashes can result in traumatic injury and death,with motor vehicle injuries accounting for a vast proportion of fatal injuries.Due to the lack of traffic accidents big data,there are only a few studies which investigatesaccidents involving motor vehicles in China,and research related to drivers’ injury severity analysis is also lacking.The fixed parameter Logit models including the ordered logit(OL)model,multinomial Logit(MNL)model,generalized ordered Logit(GOL)model,as well as the partial proportional odds(PPO)modelhave been widely appliedto investigate drivers’ injury severity.However,theaforementioned models could not account for the unobserved heterogeneity.Thus,based on the traffic accidents data obtained from the Bureau of Traffic Management at the Ministry of Public Securityin a city from China,7525 accidents involving two motor vehicles are selected from the database and used for analysis.In this study,the injury severity of two drivers in an accident who suffered from the most injuryis investigated andisconsidered as the dependent variable.Thirty potential factors are selected from four aspects involving human elements,vehicle characteristics,road features and environmental conditions,and are regarded as the independent variables.Thefixed parameter Logit model as well as the random parameter Logit(RPL)model are developed to identify risk factors influencing drivers’ injury severity in crashes,and thesemodels’ fitness as well as the prediction accuracy are compared.Then,the elasticity analysis is appliedto assess the effect of significant independent variables on drivers’ injury severity.Moreover,strategies that may reduce drivers’ injury severity in accidents are proposed based on the results revealed in this study.Comparasion results of the fixed and random parameter logit model reveal that the OL model violates the parallel-lines assumption,while the MNL modelviolates the IIA assumptions,the average prediction accuracy of GOL modelis 70.007%,which is the highest among the fixed parameter logit models.While,for the PPO model the valus of AIC and BIC are smallest.However,the aforementioned models could not account for the unobserved heterogeneity.While the RPLmodelcan account for the unobserverd heterogeneity,and there were five random parameters,which are normally distributed.Besides,values for the goodness of fit and prediction accuracy for the RPL modelare higher than that with the fiexd parameter logit model,which means that the RPL modelouterperforms the fixed parameter logit model.Based on the results revealed inthis research,this study proposes some targeted strategies from three aspects involving safety education,traffic management and enforcement as well as engineering.The founding of this research provides a theoretical basis for formulating countermeasures to reduce injury severity in accidents as well as improving traffic safety level.