Objective:
In poorer communities, smoking has demonstrated as an indicator of social inequalities in mortality. Adolescents from poor family are often exposed to smoking and are more vulnerable to smoke. This study quantified 1) the association between family characteristics and adolescent smoking in Indonesia and 2) the mediating role of the family smoking environment in this association.

Methods:
A cross-sectional survey was conducted in eight Indonesian cities among 2,661 students aged 13 to 18 years old. Multilevel logistic regression was used to estimate the association between family characteristics (parent’s educational attainment, wealth, and religion) and adolescent smoking. Mediation analysis using Generalized Structural Equation Modelling (GSEM) quantified how much the family smoking environment (family smoking status, parental control, and parental permissiveness of smoking) mediated these associations. Analyses were stratified by gender.

Results:
Smoking prevalence was 53.2% among boys and 7.7% among girls. The correlation between smoking and wealth were not significant in boys (wealthier vs poorer: OR 0.97, 95%CI 0.68-1.39), but was significant in girls (wealthier vs poorer: OR 0.49, 95%CI 0.27 – 0.89). Smoking was less favourable among Christian boys (OR 0.56, 95%CI 0.35-0.89) and Buddhist-Hindu girls (OR 0.23, 95%CI 0.05 – 0.97) compared to Muslims. Parental education was an insignificant predictor of smoking in boys and girls. Smoking environment significantly mediated the association between smoking in boys and parental education (moderate: 69.9%; low: 88.7%) and between smoking and religions in both genders (Christian boys: 46.9%; Buddhist-Hindu girls: 297%). Significant mediation indicates parental control and parental permissiveness as mediators.

Conclusions:
The current study demonstrates that social inequality in smoking remains high in girls and smoking environment is important in social patterning of adolescent smoking. Predictors of and mediation effect in smoking among girls and boys differed, hence, future tobacco control campaign targeting adolescents may not necessarily identical for both groups.