High adolescent alcohol consumption is predictive for alcohol problems later in life. To tailor interventions, early identification of risk groups for adolescent alcohol consumption is important. The IMAGEN dataset was utilized to investigate predictors for problematic alcohol consumption at age 18-20 years as a function self and parental personality and drug-related measures as well as life-events and cognitive variables all assessed at age 14 years (N = 1404). For this purpose the binary partitioning algorithm ctree was used in an explorative analysis. The algorithm recursively selects significant input variables and splits the outcome variable based on these, yielding a conditional inference tree. Four significant split variables, namely Place of residence, the Disorganization subscale of the Temperament and Character Inventory, Sex, and the Sexuality subscale of the life-events questionnaire were found to distinguish between adolescents scoring high or low on the Alcohol Use Disorders Identification Test about five years later (all p < 0.001). The analyis adds to the literature on predictors of adolescent drinking problems using a large European sample. The identified split variables could easily be collected in community samples. If their validity is proven in independent samples, they could facilitate intervention studies in the field of adolescent alcohol prevention.
Keywords: Adolescence; Alcohol consumption; Conditional inference trees (ctree); Hierarchical associations.
Copyright © 2019. Published by Elsevier Ltd.