Transition to college: A classification and regression tree (CART) analysis of natural reduction of binge drinking

Approximately one in five teens that drank heavily in high school reduces or discontinues consumption while in college. Multiple paths might lead to the common outcome of natural reduction in heavy drinking. Statistical modeling of this complex process of natural reduction is a challenge with standard linear statistics. The purpose of this paper is to use a new statistical procedure, Classification and Regression Tree (CART), to model the equifinality of reduction in drinking by college students who drank heavily as adolescents. An appealing aspect of CART is that the resulting tree model that can easily be interpreted and applied by those who work with adolescents during the important transition from high school to college. Of 201 college students who first binged on alcohol while in high school, 71 (35.3%) denied heavy or binge drinking within the previous three months (Natural Reducers). The final model accurately classified 84.6% of the students as either continued heavy drinkers or natural reducers. Sensitivity was modest (accurate identification of 67.6% of the reducers); however, specificity was strong (correct classification of 93.8% of the continued heavy drinkers). The model revealed four pathways to natural reduction in drinking. Predominant in each path was the influence of social factors that maintain continued drinking (e.g., social facilitation outcome expectancies, perception of friends' drinking) or facilitate natural reduction (e.g., regular church attendance). The results support the application of CART to model health behaviors across the transition from adolescence to young adulthood.