The temporal stability of factors affecting driver injury severity in run-off -road crashes: A random parameters ordered probit model with heterogeneity in the means approach

This study presents an effort to investigate the determinants of driver injury severity in run-off-road (ROR) crashes. In order to account for unobserved heterogeneity, a random parameter ordered probit with heterogeneity in the means approach is applied. The police-reported ROR crash data that occurred from 2014 to 2017 in the state of North Carolina is used. Four injury-severity levels are defined: property damage only (PDO), possible injury (PI), non-incapacitating (N), and F/I (by merging fatal and incapacitating). The driver, crash, roadway, environmental characteristics that potentially affect the driver injury severity are explored. Besides, the temporal stability of modeling results among the four-time periods is investigated using a series of likelihood ratio tests. Significant temporal instability is found, indicating underestimating the temporal instability might result in unreliable conclusions. Estimation results demonstrate that the indicators, including male driver, alcohol, and curved roadways, increase the possibility of fatal and incapacitating injuries in the ROR crash in most of the year periods.