Abstract

OBJECTIVE: To test whether baseline data from a randomised clinical trial are predictive of initiation of tobacco use over a two-year follow-up interval, and to discuss results in the context of a theoretical model. DESIGN: Secondary, non-experimental analyses of data collected from a prospective cluster-randomised clinical trial comparing an intervention with a control condition for reduction of tobacco incidence rates. Orthodontic offices in southern California were recruited and randomised to an experimental or control group. Patient participants were sampled within each office, and completed a short survey, repeated two years later. SUBJECTS: 13,923 patients, 11-18 years of age, randomly sampled from each office. MAIN OUTCOME MEASURES: The ability of baseline data to predict initiation of tobacco use over the two-year follow-up interval was tested through a series of logistic regression models. Significant predictors and their interactions were identified in fixed-effects models, and verified in a mixed-effects logistic regression model to account for cluster randomisation. RESULTS: Clinician advice against tobacco use was associated with a lower rate of tobacco use initiation among young people whose peer group considered smoking socially desirable. Rates of initiation increased with age, but this association differed by gender and by whether the adolescent had been offered tobacco within 30 days prior to the baseline assessment. People from minority groups were less likely to initiate tobacco use than whites, and young people engaging in other risk practices were more likely to initiate tobacco use. CONCLUSIONS: Findings support predictions based on learning theory that social processes are critical in the development of health-risk behaviours. Future preventive efforts should target changing the density with which young people encounter pro- and anti-tobacco prompts and consequences in the community.