The interactive effect on injury severity of driver-vehicle units in two-vehicle crashes
Introduction
Due to the enormous losses resulting from highway traffic crashes, reducing the severity of injury sustained by crash-involved roadway users has long been a prominent concern of transportation agencies and vehicle manufacturers (Savolainen, Mannering, Lord, & Quddus, 2011). In the field of highway safety research, considerable efforts have also been devoted to gain a better understanding of how the relevant factors affect the degree of injury severity, which is expected to provide useful suggestions for laws, regulations and countermeasures aimed at mitigating crash severity. According to the extant traffic safety research literature, the contributing factors to injury severity refer to the driver's demographic or behavioral characteristics, the vehicle's technical characteristics, roadway conditions and environmental factors at the time and place of crash occurrence. Because crash injury severity is generally categorized into discrete levels, in order to model its relationship with the risk factors, logit and probit models and a number of their variations have been proposed, ranging from ordered models (Khattak, Kantor, & Council, 1998), Bayesian hierarchical (Huang, Chin, & Haque, 2008)/simultaneous (Ouyang, Shankar, & Yamamoto, 2002) models, bivariate (Lee & Abdel-Aty, 2008)/multivariate (Winston, Maheshri, & Mannering, 2006) models, and nested logit model (Shankar & Mannering, 1996), to random parameter model (Milton, Shankar, & Mannering, 2008), Markov switching multinomial model (Malyshkina & Mannering, 2009) and their mixed versions (Eluru and Bhat, 2007, Huang et al., 2011, Zoi et al., 2010). Besides, some data mining techniques, such as the neural network (Abdelwahab and Abdel-Aty, 2001, Abdelwahab and Abdel-Aty, 2002, Chimba and Sando, 2009, Delen et al., 2006, Zeng and Huang, 2014b, Zeng et al., 2016), classification and regression tree (Chang & Wang, 2006), and support vector machine (Li, Liu, Wang, & Xu, 2012) have also been used as they exhibit better nonlinear approximation performance than traditional discrete outcome models (Please refer to Savolainen et al. (2011) and Mannering and Bhat (2014) for more detailed descriptions and assessments of these models). Although the analytical methods have been advanced continually in the past decades, when the injury severity of each driver-vehicle unit in two-/multi-vehicle crashes is analyzed, the driver-specific and vehicle-specific factors of only itself are considered at most cases. The impact of the counterpart involved in the same crash on the injury degree of the subject driver-vehicle unit has rarely been explored.
Recently, a quantity of studies has been focused on crash compatibility (Fredette et al., 2008, Huang et al., 2014a, Huang et al., 2014b, Huang et al., 2016, Huang et al., 2011, Toy and Hammitt, 2003, Wenzel and Ross, 2005), which is deemed as a critical criterion for evaluating a vehicle's safety performance. In these studies, the variations in vehicle incompatibility (such as pickups versus cars) were found to significantly affect the injury severity outcomes of all drivers when a crash occurs. According to Hollowell and Gabler (1996), a vehicle's incompatibility is the combination of its crash worthiness (i.e. the self-protective capacity) and its crash aggressivity (i.e. the hazardousness imposed on the other vehicle(s)) in the collision. Specifically, the vehicle incompatibility refers to the differences of vehicle design with respect to mass, geometry, structure, etc. Therefore, significant distinction was found in the crash worthiness and aggressivity of different vehicle types (Fredette et al., 2008, Huang et al., 2011, Toy and Hammitt, 2003) and models (Huang et al., 2014a, Huang et al., 2014b, Huang et al., 2016). In addition to the crash aggressivity, as Toy and Hammitt (2003) argued, some driver factors of the subject vehicle may also impact the injury severity of the driver of the other vehicle with which it collides. However, only the effect of crash aggressivity was analyzed in the previous research. A plausible speculation is that the driver-vehicle units in a two-/multi-vehicle crash interact and the attributes of each driver-vehicle unit may have significant effects on the severity of the injuries incurred by all involved drivers.
In this study, to empirically analyze the interactive effect on injury severity of driver-vehicle units, a two-vehicle crash dataset was obtained from the Florida Department of Highway Safety and Motor Vehicles (DHSMV). The factors of both driver-vehicle units together with the crash configuration ones are used as the independent variable in the hierarchical ordered logit (HOL) model for analyzing the injury severity of each driver in a two-vehicle crash. The analyzed results are also expected to provide some new directions for traffic safety education, enforcement and engineering. The rest of this paper is organized as follows. First, the collected crash data are preprocessed and their statistical characteristics are presented. Then, the specific structure and estimation process of the proposed analytical method are illustrated. After that, the estimation results are discussed in details. Finally, some specific recommendations for traffic safety improvement by education, enforcement and engineering are made and directions for future research are presented.
Section snippets
Data preparation
The 2007 crash data set from the Florida DHSMV is used for the research. For convenience of investigating the interaction of driver-vehicle units, the analysis focuses on two-vehicle crashes and only the crashes with complete information on the factors as listed in Table 1 are extracted, resulting in a total of 12,834 driver-vehicle units involved in 6417 crashes.
The injury severity of each driver is used as the dependent variable, since driver injury records are generally more complete in the
Model development
Given the ordinal nature of the response variable, an ordered discrete outcome model is more appropriate than an unordered model. Moreover, in a two-vehicle collision, the injury levels of both drivers involved may be affected by certain common unobserved factors, resulting in within-crash correlation (Huang et al., 2008). The Bayesian HOL model which allows for the consideration of both ordinal response and within-crash correlation is adopted in this analysis (Huang et al., 2011).
In the
Result analysis
The results of the parameter estimation of driver-specific and vehicle-specific factors are summarized in Table 4. These results might provide with useful information about how to reduce the impact of an impending crash, which could be used to modify curriculum by traffic safety educators. Specifically, older drivers are more likely to make themselves injured but less likely to injure the drivers in the other vehicles.1
Conclusions and future research
This study empirically investigates the interactive effect on injury severity of driver-vehicle units based on a two-vehicle crash dataset in Florida, USA. A state-of-the-art method, the Bayesian HOL model, is adopted for the empirical analysis, where the driver-specific and vehicle-specific factors of the other vehicle are added into the set of explanatory variables of the subject driver injury severity, together with its own driver-vehicle attributes and the crash configuration factors.
Acknowledgments
This research was jointly supported by the Natural Science Foundation of China (nos. 51378222, 51578247, 71371192) and a grant from the Joint Research Scheme of National Natural Science Foundation of China/Research Grants Council of Hong Kong (no. 71561167001). We would like to thank Dr. Mohamed Abdel-Aty at the University of Central Florida for providing the data.
Qiang Zeng received his B.E. and Ph.D. degrees in transportation engineering from Central South University, China, in 2010 and 2015 respectively. Now, he is working as a postdoctoral fellow in the School of Civil Engineering and Transportation, South China University of Technology, China. His research interests cover traffic safety analysis, transportation system modeling and public transit operation & management.
References (39)
- et al.
Analysis of traffic injury severity: An application of non-parametric classification tree techniques
Accident Analysis and Prevention
(2006) - et al.
Identifying significant predictors of injury severity in traffic accidents using a series of artificial neural networks
Accident Analysis and Prevention
(2006) - et al.
Macroscopic hotspots identification: A Bayesian spatio-temporal interaction approach
Accident Analysis and Prevention
(2016) - et al.
A joint econometric analysis of seat belt use and crash-related injury severity
Accident Analysis and Prevention
(2007) - et al.
Safety impacts due to the incompatibility of SUVs, minivans, and pickup trucks in two-vehicle collisions
Accident Analysis and Prevention
(2008) - et al.
Indexing crash worthiness and crash aggressivity by major car brands
Safety Science
(2014) - et al.
Severity of driver injury and vehicle damage in traffic crashes at intersections: A Bayesian hierarchical analysis
Accident Analysis and Prevention
(2008) - et al.
Crash protectiveness to occupant injury and vehicle damage: An investigation on major car brands
Accident Analysis and Prevention
(2016) - et al.
Indexing crash worthiness and crash aggressivity by vehicle type
Accident Analysis and Prevention
(2011) - et al.
Effects of work zone presence on injury and non-injury crashes
Accident Analysis & Prevention
(2002)
Presence of passengers: Does it increase or reduce driver's crash potential?
Accident Analysis and Prevention
Using support vector machine models for crash injury severity analysis
Accident Analysis and Prevention
Markov switching multinomial logit model: An application to accident-injury severities
Accident Analysis and Prevention
Analytic methods in accident research: Methodological frontier and future directions
Analytic Methods in Accident Research
Highway accident severities and the mixed logit model: An exploratory empirical analysis
Accident Analysis and Prevention
The statistical analysis of highway crash-injury severities: A review and assessment of methodological alternatives
Accident Analysis and Prevention
An exploratory multinomial logit analysis of single-vehicle motorcycle accident severity
Journal of Safety Research
Drug use and the severity of a traffic accident
Accident Analysis and Prevention
Road network safety evaluation using Bayesian hierarchical joint model
Accident Analysis and Prevention
Cited by (38)
Machine learning-based injury severity prediction of level 1 trauma center enrolled patients associated with car-to-car crashes in Korea
2023, Computers in Biology and MedicineCitation Excerpt :Unlike fixed-material collisions, vehicle incompatibilities (e.g., passenger cars versus SUVs) in C2C crashes have contributed to injury severity outcomes [16,62–64]. These vehicle mismatches of body structure increased the risk of injury severity to MVOs with disadvantageous self-protective capacities due to vehicle differences, such as mass, weight, geometry, and stiffness, based on Newtonian mechanics [11,14–16,51,54,65]. Zeng et al. (2016) reported that vans and trucks had stronger self-protection and aggressivity than passenger vehicles [16].
Jointly analyzing freeway primary and secondary crash severity using a copula-based approach
2023, Accident Analysis and PreventionTemporal stability of factors affecting injury severity in rear-end and non-rear-end crashes: A random parameter approach with heterogeneity in means and variances
2022, Analytic Methods in Accident ResearchCitation Excerpt :The model estimation results indicate that the AADT indicator plays a critical role in all six models, with the stable positive marginal effects yielding increased occurrence of no injury rear-end and non-rear-end crashes over time. This may be due to the lower travel speeds associated with high traffic volumes (Christoforou et al., 2010; Zeng et al., 2019), whereas a vehicle traveling at a high speed increases the severity outcome of any crash involving it (Zeng et al., 2016). It is worth noting that the AADT is positively related to severe and fatal injury rear-end crashes but negatively associated with severe and fatal injury non-rear-end crashes across the three years, revealing its potential heterogeneity, which may be due to the different generation mechanisms of rear-end and non-rear-end crashes.
Qiang Zeng received his B.E. and Ph.D. degrees in transportation engineering from Central South University, China, in 2010 and 2015 respectively. Now, he is working as a postdoctoral fellow in the School of Civil Engineering and Transportation, South China University of Technology, China. His research interests cover traffic safety analysis, transportation system modeling and public transit operation & management.
Huiying Wen received her Ph.D. degree in transportation engineering from South China University of Technology, China, in 2007. She is a professor, a deputy dean of School of Civil Engineering and Transportation, South China University of Technology. Her research interests refer to traffic safety analysis, transportation planning and management.
Helai Huang received his B.E. degree in civil engineering and M.E. degree in hydraulic engineering from Tianjin University, China, in 2000 and 2003 respectively; and Ph.D. degree in transportation engineering from National University of Singapore, Singapore in 2007. He worked as a research fellow in National University of Singapore during 2007–2008 and a research associate & graduate faculty scholar in University of Central Florida, USA, during 2008–2010. In December, 2010, he joined School of Traffic and Transportation Engineering, Central South University. Currently, he is a professor, the director of Urban Transit Research Center and the deputy dean of the school. He is also an associate editor of Accident Analysis & Prevention, an editorial member of several academic journals such as Analytic Methods in Accident Research and Journal of Geography and Regional Planning, and a committee member of Transportation Research Board ABJ80: Statistical Methods.