Using a classification tree modeling approach to predict cigarette use among adolescents in the United States

Abstract Objective: The purpose of this study was to screen pertinent variables to identify ordered relations that provide easily interpretable and accurate predictions of the probability of cigarette use among adolescents using a classification tree modeling approach. Methods: This cross-sectional study included a national sample of 3717 U.S. adolescents aged between 12 and 20 years old from the 2016 National Survey on Drug Use and Health. Results: The results indicated that age was the most influential variable, followed by depression, race/ethnicity, family income, gender, and alcohol abuse and dependence. Additionally, several interaction emerged that identified higher and lower cigarette use profiles: youth who were between 18 and 20 years old and self-identified as non-Hispanic White, Native American/Alaska Native, and “Other” racial/ethnic group and African American, Asian, and Latinx adolescents, with depressive symptoms were at higher risk of cigarette use; while youth who reported lower family incomes, were 16–17 years old, who identified as African American, Asian, and Latinx, were also likely to report lower use of cigarettes when they reported lower depressive symptom scores. Discussion: These results are discussed relative to practice implications.