Accelerationbased occupant injury criteria for predicting injury in realworld crashes.

This paper presents an estimate of the probability of serious occupant injury in frontal crashes based on two vehicle acceleration-based metrics: the 10 ms peak acceleration and the 50 ms peak acceleration. Both of these metrics are used to evaluate injury potential in crash test involving roadside hardware, such as guardrail. For this study, Event Data Recorder (EDR) data provides vehicle kinematics information for real world crashes with known injury outcomes. Based on a data set of 180 cases, binary logistic regression was used to generate injury risk curves for belted and unbelted occupant data subsets. Model fit statistics and a Receiver Operator Characteristic (ROC) analysis was then used to compare the injury predictive capability of these two metrics. No statistically significant difference was found between the injury prediction capabilities of the 10 ms and 50 ms peak acceleration metrics.