Using Ordered Modeling to Identify the Most Significant Factors that Increase the Severity of Single-Vehicle Collisions
Author(s): Dabbour, Aly
Slidedeck Presentation:
Abstract:
Approximately 60% of fatal traffic collisions are related to drivers who lose control over their own vehicles with no other vehicles involved (single-vehicle collisions). Several research studies attempted to identify the factors that increase injury severity related to a single-vehicle collision given that such a collision has already occurred. Most of those studies were based on compiling data from different years into a single dataset that was used for analysis. However, the factors that affect the severity of single-vehicle collisions might be temporally unstable so that they might have been significant during the period being analyzed and then their significance might have changed due to the ongoing changes in vehicle technologies, drivers' attitudes, traffic volumes, road conditions, law-enforcement technologies, and implemented policies. Understanding the temporal stability trends of the factors affecting injury severity would help researchers assessing the effectiveness of implementing different safety treatments so that they could identify whether any safety improvements are attributed to the specific treatment or simply attributed to the temporal instability of the factors being addressed. The aim of the study is to investigate the temporal stability of the factors that increase drivers' injury severity in single-vehicle collisions. The findings of the study will assist decision makers identifying the most temporally-stable factors that increase the severity of drivers' injuries so that different resources may be allocated to reduce the impacts of those factors. The findings of this study will also provide quantitative measures that may be used to determine the feasibility of implementing different countermeasures in reducing the severity of drivers' injuries related to single-vehicle collisions. The study method is to analyze all single-vehicle collisions that occurred in North Carolina from January 1, 2007 to December 31, 2013. A separate ordered regression model will be developed for each analysis year, and the models are compared together to determine whether the effects of the identified factors are temporally stable throughout the analysis period. To further highlight the scientific contributions of the study, an overall model will also be developed for the overall analysis period so that a comparison may be made between the limited insights provided by the overall model and the in-depth insights provided by studying the year-to-year changes in the factors being investigated. Several factors that had been found in other research studies to increase the severity of drivers' injuries are found in this study to be temporally unstable and therefore their effects might have been significant during a certain period and then became insignificant afterward. The factors that were found to be temporally unstable include driving on an undivided highway, lighting conditions, season of the year, and being a senior driver (above 65 years). Several other factors were found to be temporally stable regarding their effects on the severity of drivers' injuries in single-vehicle collisions. Those temporally-stable factors include favorable weather conditions, driver's gender, exceeding the speed limit, failing to properly use seat belt, and driving under the influence of alcohol, if the collision occurred on a highway curve, and if the collision occurred on a rural highway. If safety treatments were to be applied by compiling data from different years into one dataset (similar to what was done in most previous research studies) without testing the temporal stability of the factors being addressed (by testing year-to-year variations within the data), it would be difficult to determine whether any potential safety improvements (in terms of reduced severity of drivers' injuries) are attributed to the applied safety treatments or to the temporal instability of the identified factors.