Traffic crash fatalities in rural areas account for more than
50% of all reported fatalities in traffic crashes, and most of them
occur at rural non-interstate locations. Therefore, it is of practical
significance to examine the injury patterns and the contributing
factors related to injury severity outcomes in rural non-interstate
crashes. In this paper, a hierarchical ordinal logit model incorporating
between-crash variance is utilized to capture the ordinal
nature of driver injury severities and identify the significant factors
for injury severity prediction. FB inference is applied to assess
the influence of these factors, including crash features, environmental
conditions, vehicle features and driver characteristics. The
95% BCI is utilized for variable significance tests. Without considering
the between-crash variance, an ordinary ordered logit model
was examined for model comparison purpose. The research results
illustrate that the proposed model structure is better in analyzing
the selected dataset, according to the DIC model performance
measurement, and therefore it is necessary to take into account the
between-crash variance in driver injury severity modeling.
Ten variables regarding crash, vehicle and driver characteristics
are identified to be significant in driver injury severity prediction
in rural non-interstate crashes. In this analysis, road segments far
from intersections, wet road surface conditions, and driver seatbelt
use tend to reduce driver injury severity levels. Single-vehicle
crashes, multiple-vehicle crashes, severe vehicle damage in a crash
and driver with alcohol or drug impairment increase the potential
of higher driver injury severities and fatalities. As to collision typ