MODELS FOR THE PREDICTION OF SEVERE INJURY

The data from the National Crash Severity Study (NCSS) were used to investigate the relationship between occupant injury severity and associated crash conditions. Models that predict an individual's probability of severe injury are presented for side and front impacts. The most important variables in predicting injury severity were found to be Delta V (a measure of crash severity) and occupant age. Although the overall predictive capability of the injury severity models was good, the proportion of correct prediction of cases with severe injury was low. Model mispredictions were analyzed to characterize specific injuries for which the assigned levels of severity were inconsistent with the models.