Analysis of Fatal Run-Off-the-Road Crashes Involving Overcorrection

In an attempt to identify characteristics that have a strong positive association with overcorrection, data on 579 fatal run-off-the-road (ROR) crashes on state roadways in Florida were analyzed with logistic regression techniques. To overcome shortcomings of traditional analysis methods relying primarily on crash reports, this study relied on case reviews using a broad variety of resources from various disciplines. The data set in this study represents a significant enhancement in accuracy and completeness over that in the initial crash reports; overcorrection was identified using traffic homicide investigation reports. A full model involving 23 explanatory variables was developed, and backward stepwise regression was conducted to identify the most predictive variables. Overcorrection cases were strongly associated with alcohol, inattention, high speed, and fatigue and sleeping. Outcomes ranged from roadway departure in the opposite direction, to vehicle impact on returning to the roadway. The study indicated that females were approximately 40% more likely to overcorrect in a fatal ROR crash than males were, with-the greatest disparity occurring among middle-aged drivers. Further, while fewer than 20% of fatal ROR crashes occurred where rumble strips were present, drivers were more than 50% more likely to overcorrect than when they were not present. On high-speed (70 mph) roadways with rumble strips, there was almost an 80% higher risk of overcorrection in the crash. Thus, while it appears that rumble strips are effective in preventing many ROR crashes, the contribution of auditory and vibratory sensations of rumble strips to panic oversteering should also be investigated.