This technical report, a companion to the volume ii user manual by the same name, describes the development and testing of predictive models for identifying individuals with a high risk of alcohol/related (a/r) crash involvement. From a literature review and an analysis of north carolina accident data, six groups of drivers at high risk of a/r crashes were identified: 16 to 20-year-old males; 21 to 24-year-old males; persons with previous dui convictions; persons with three or more moving violations; persons recently divorced; and persons recently released from prison. Using north carolina data through 1974, predictive models were developed for each of these groups to predict 1975 a/r accident involvement proportions for subgroups within each high-risk group. Prospective analyses of the models' predictive capabilities using 1976 crash data indicated that, in general, they effectively identify the driver subgroups that have the highest risk of a/r crash involvement. Because most of the information used in the model development is readily available in most governmental jurisdictions, the models can be adapted for use in areas other than north carolina. The primary drawback to widespread use of the models is the lack of demonstrably effective countermeasure programs that significantly reduce the rate of a/r crash involvement for the identified driver subgroups.
Abstract