Individual-Level Predicted Probabilities From Discrete Choice Experiments Predict Subsequent Tobacco Product Initiation Among Mexican Adolescents
Desiree Vidana-Perez, Liyan Xiong, Inti Barrientos-Gutiérrez, Diego F. Leal, James Hardin, James F. ThrasherPURPOSE
Discrete choice experiments (DCEs) are increasingly used to evaluate adolescent and adult perceptions of and responses to pro-tobacco marketing and anti-tobacco interventions; however, this area of research is limited due to its cross-sectional nature. This study assessed whether individual-level predicted probabilities derived from DCEs can predict subsequent nicotine product use among Mexican adolescents.
METHODS
In a cohort of students in 20 high schools in two Mexican cities (Guadalajara and Tijuana), we analyzed data from analytic samples of students who were followed up six months later and, at baseline reported they never used: (1) cigarettes (n = 3,814); (2) e-cigarettes (n = 3,034). Students completed a 2 × 2 × 4 within-subject DCE with the same eight choice sets, each containing cigarettes and e-cigarettes with different flavors (mango, mint, tobacco, or berries) and brands (two types). Individual-level predicted probabilities for interest in trying cigarettes, e-cigarettes or neither were estimated using Hierarchical Bayes models. Aggregated probabilities were used as primary independent variables in subsequent logistic regression models assessing associations between predicted probabilities and product initiation at wave 2. Separate unadjusted and adjusted logistic models regressed initiation of cigarettes and e-cigarettes at follow-up on DCE-derived predicted probabilities, adjusting for socio-demographics, susceptibility to use either product, peer use of either product, and sensation seeking.
RESULTS
At baseline, individual-level predicted probabilities of choosing e-cigarettes were higher than those for cigarettes (25.2% and 2.9%, respectively), as was initiation of e-cigarette use at 6-month follow-up (3.8% and 7.1%, respectively). In adjusted logistic regression models, higher baseline predicted probabilities for interest in trying either product was positively associated with initiation (cigarettes: OR = 1.27; e-cigarettes: OR = 1.16,
CONCLUSION
Predicted probabilities for interest in trying cigarettes and e-cigarettes that were derived from DCEs had strong associations with initiation of nicotine product use among Mexican adolescents, including after adjustment for traditional risk factors. Evidence for their predictive validity supports the integration of DCEs into research on product-specific appeal in future tobacco control research, including policy evaluation and simulation studies.