Aggression and delinquent behavior in a large representative sample of high school students: Cannabis use and...

Since conduct problems (CP) vary distinctly across youths, better subtyping CP may be an important vehicle to study specific risk factors associated to differential patterns of CP. In a sample of 63,196 adolescents, we employed a two-step method to the identify such CP patterns and to help classify youths based on several sociodemographic and psychopathological risk factors associated with CP.

K-means clustering methods were first used to reduce the heterogeneity of CP by analyzing patterns of aggressive (AGG) and rule-breaking (RB) behaviors. A multi-class Classification and Regression Tree approach was further employed to examine the hierarchical interactions between risk factors specific to the emergence of different CP patterns.

Results revealed a three-cluster solution: (i) Low AGG-RB, (ii) High AGG and low RB, and (iii) High AGG-RB. The frequency of cannabis use, level of victimization and hyperactivity symptoms were the three factors best discriminating youths' membership to distinct patterns of CP. The model displayed a moderate to strong discriminatory capacity.

Although this study provides evidence of key factors that may increase the risk of youths following specific patterns of disruptive behavior, additional research is necessary to clarify the etiology, longitudinal trajectories and outcomes related to these patterns.