Background and aims Conduct Problems (CP) can differ in terms of continuity/desistence across different stages of life such as childhood and adolescence. Although there have been a number of research efforts to investigate individual-level factors associated with continuity/desistence of CP in adolescence, less is known about school factors that may potentially underlie continuity/desistence of CP such as school-level factors. Our aim is to systematically investigate a number of school-level factors that may predict CP trajectories while controlling for individual-level factors.

Methods We run longitudinal latent class analysis (LLCA) to identify trajectories of CP across adolescence using the Learning Together study dataset. We then investigated the role of a number of school-level factors in predicting class membership using multinomial logistic regression. We identified two classes of CP: a stable low and a moderate-high class across males and females.

Results A number of school-level factors such as student-teachers’ relationships, sense of belonging to the school and participation in school activities predicted persistent patterns of CP in males and females separately. Positive school atmosphere was found to be strongly associated with a lower risk of persistent CP across males and females.

Conclusions Student-teachers’ relationships, sense of belonging to the school and participation in school activities are important aspects that researchers and schools should consider when implementing prevention and intervention programs for youth with CP and antisocial behaviour.