Objective: This study investigated the confounding effects of factors on injury outcomes for all occupants in fatal single-vehicle crashes that involved a rollover event. Method: A generalized ordered logit model was used to investigate the role of roadway attributes, environmental factors, driver characteristics, and vehicle features in injury severity outcomes for occupants. Five years of single-vehicle rollover crash data for the United States were studied. Results: Results showed that the likelihood of serious and fatal injuries increases in rollover crashes with partial or complete ejection of the occupant, no seat belt use, speeding, higher posted speed limits, roadside and median rollovers, undulating terrain, blacktop road surface, and rural roads. We also found that evening, weekdays, previous driver crash, careless or inattentive driving, driver–passenger engagement, aggressive driving, and vehicle type affect injury severity. The deployment of airbags was associated with fewer serious and fatal injuries. Regional differences were found for injury severity outcomes in rollover crashes. Conclusions: The study provides valuable insight for reducing injury severity in single-vehicle crashes where a rollover occurs. Several proven countermeasures may prevent rollovers or reduce injury severity. These strategies include increasing seatbelt use, posting lower speed limits and installing speed enforcement cameras in high-risk areas, flattening roadside embankments, and promoting in-vehicle stability enhancement systems such as electronic stability control and rollover-activated side curtain airbags.

Safety at highway rail grade crossings (HRCs) continues to be a serious concern despite improved safety practices. Accident frequencies remain high despite increasing emphasis on HRCs safety. Consequently, there is a need to re-examine both the design practices and the safety evaluation methods at HRCs. Previous studies developed accident prediction models by incorporating highway, crossing inventory, rail, and vehicle traffic characteristics, but none of these factors considered population in the vicinity of HRCs. This study developed a binary logit regression model to predict accident likelihood at HRCs by incorporating various contributory factors in addition to population (based on census blocks 2010) within five miles of crossings. Previous North Dakota accident data from 2000 to 2016 was analyzed and used in the model development. The model results show that the number of daily trains, the maximum typical train speed, the number of through railroad tracks, and the number of highway/traffic lanes all affect accident likelihood. The presence of pavement markings in the form of stop lines helps reduce accident probability, while populations within five miles of HRCs have a positive relationship with crash likelihood. This study will help transportation agencies improve HRC safety.