Search

• Crash prediction by injury severity and vehicle damage was conducted and compared. • Multivariate Poisson-Lognormal (MVPLN) models were estimated. • Joint Negative Binomial-Generalized Ordered Probit Fractional Split (NB-GOPFS) models were estimated. • Unobserved heterogeneity was accommodated by both models. • A surrogate measure for injury severity was evaluated. Introduction: Predicting crash counts by severity plays a dominant role in identifying roadway sites that experience overrepresented crashes, or an increase in the potential for crashes with higher severity levels. Valid and reliable methodologies for predicting highway accidents by severity are necessary in assessing contributing factors to severe highway crashes, and assisting the practitioners in allocating safety improvement resources. Methods: This paper uses urban and suburban intersection data in Connecticut, along with two sophisticated modeling approaches, i.e. a Multivariate Poisson-Lognormal (MVPLN) model and a Joint Negative Binomial-Generalized Ordered Probit Fractional Split (NB-GOPFS) model to assess the methodological rationality and accuracy by accommodating for the unobserved factors in predicting crash counts by severity level. Furthermore, crash prediction models based on vehicle damage level are estimated using the same two methodologies to supplement the injury severity in estimating crashes by severity when the sample mean of severe injury crashes (e.g., fatal crashes) is very low. Results: The model estimation results highlight the presence of correlations of crash counts among severity levels, as well as the crash counts in total and crash proportions by different severity levels. A comparison of results indicates that injury severity and vehicle damage are highly consistent. Conclusions: Crash severity counts are significantly correlated and should be accommodated in crash prediction models. Practical application: The findings of this research could help select sound and reliable methodologies for predicting highway accidents by injury severity. When crash data samples have challenges associated with the low observed sampling rates for severe injury crashes, this research also confirmed that vehicle damage can be appropriate as an alternative to injury severity in crash prediction by severity. [ABSTRACT FROM AUTHOR]

Abstract: This paper introduces dynamic time series modeling in a Bayesian framework to uncover temporal patterns in highway crashes in Connecticut. Existing state sources provide data describing the time for each crash and demographic attributes of persons involved over the time period from January 1995 to December 2009 as well as the traffic volumes and the characteristics of the roads on which these crashes occurred. Induced exposure techniques are used to estimate the exposure for senior and non-senior drivers by road access type (limited access and surface roads) and area type (urban or rural). We show that these dynamic models fit the data better than the usual GLM framework while also permitting discovery of temporal trends in the estimation of parameters, and that computational difficulties arising from Markov Chain Monte Carlo (MCMC) techniques can be handled by the innovative Integrated Nested Laplace Approximations (INLA). Using these techniques we find that while overall safety is increasing over time, the level of safety for senior drivers has remained more stagnant than for non-senior drivers, particularly on rural limited access roads. The greatest opportunity for improvement of safety for senior drivers is on rural surface roads. [Copyright &y& Elsevier]