Characterizing patterns of smoking initiation in adolescence: comparison of methods for dealing with missing data.

INTRODUCTION: Tobacco use is common and remains one of the leading causes of preventable death in developed countries. Smoking commonly begins in adolescence, and hence, it is important to understand how smoking behavior develops during this period.
METHODS: In a U.K.-based birth cohort, we analyzed repeated measures of smoking frequency in a sample of 7,322 young adolescents. Latent class analysis was used to summarize the data, and the resulting classes of behavior were related to a range of smoking risk factors. Results from a complete case analysis were compared with estimation using full-information maximum likelihood (FIML) and estimation using multiple imputation (MI).
RESULTS: Fifty-three percent of the sample reported having smoked a whole cigarette by age 16 years. The longitudinal data were summarized by 4 distinct patterns of smoking initiation: nonsmokers (79.7%), experimenters (10.3%), late-onset regular smokers (5.5%), and early-onset regular smokers (4.5%). Social disadvantage, other substance use, conduct problems, and female sex were strongly related to being a regular smoker; however, no risk factors studied showed any strong or consistent association with experimentation. In the complete case sample, smoking prevalence was lower, and in addition, the association between different smoking patterns and covariates was often inconsistent with those obtained through FIML/MI.
CONCLUSIONS: Most young people have experimented with tobacco smoking by age 16 years, and regular smoking is established in a substantial minority characterized by social disadvantage, other substance, use and conduct disorder. Prevention strategies should focus on this subgroup as most children who experiment with tobacco do not progress to regular smoking.