Predicting survival of victims of motor vehicle crashes in New York state.

This study assesses the relative ability of three different models to predict in-hospital mortality for victims of motor vehicle crashes. The first two models, the trauma and injury severity score (TRISS) and a severity characterization of trauma (ASCOT), are models that have been used in many earlier studies and have been quoted extensively in the literature. The third model, which is developed in this study, uses essentially the same risk factors as the other two studies, but employs them in a different manner. In order to provide a fair comparison, new (logistic regression) model coefficients are fit to the first two models using the study data. The models are compared with respect to typical criteria for assessing the fit of logistic regression models as well as their ability to predict mortality for various subsets of seriously injured patients. The study concludes that the new model provides a substantially more accurate prediction of mortality, and that it may be wise for regions attempting to assess relative outcomes in their subregions to develop statistical models that are tailored to their own patients.