School context variables and students' smoking. Testing a mediation model through multilevel analysis

Abstract

Background/aims: This study applied a multilevel approach to examine the associations between school smoking policy and student smoking. It was tested whether individual characteristics are mediators of school policy effects.

Methods: On the basis of cross-sectional data from 3,364 students and school principals from 40 schools in Germany, two multilevel nonlinear regression models were computed for current smoking.

Results: In the first model, controlling for individual factors not influenced by school, smoking bans for students and evidence-based prevention activities were negatively associated with smoking prevalence. The second model included student characteristics potentially influenced by school (e.g. school engagement, peer smoking). As school variables remained significant, these characteristics obviously do not mediate school context effects.

Conclusion: School context variables influence students' smoking independent of individual factors. The data suggest that individual variables rather mediate family influences.