The URGENCY algorithm offers a means to determine the likelihood of serious injury in the event of a motor vehicle crash. The algorithm, based on historical crash data, relates crash characteristics to injury severity outcomes so that the most appropriate level of rescue care can be provided rapidly, accurately, and remotely in future systems. This article describes the development and enhancement of previous versions of the URGENCY algorithm to produce URGENCY 2.1. This algorithm allows for crash data, including crash characteristics, occupant characteristics, as well as vehicle factors useful for crash severity prediction. Recent enhancements of the URGENCY algorithm include an enhanced approach to the prediction of injury in the event of multiple impacts, improved criteria to classify crashes by mode, and the inclusion of additional data to train predictive models. Following enhancement, each algorithm has been tested for it's predictive value based on a sample of evaluation cases not used for model development. The results of this evaluation are presented.
Keywords:
automatic crash notification (ACN); injury prediction; logistic regression