Exploring a diverse set of specifications related to associations between adolescent smoking, vaping, and emotional problems: a multiverse analysis.
Journal Articles
Overview
Research
Identity
Additional Document Info
View All
Overview
abstract
PURPOSE: The epidemiological landscape of adolescent smoking, vaping, and emotional problems has drastically changed over the past two decades. Whether and why these problems co-occur remains unclear, though this understanding is crucial for global policy and prevention efforts. The nature of co-occurring problems may be influenced by different researcher decisions when defining, operationalizing, and modeling these relationships. This study uses multiverse analysis (also known as specification curve analysis or vibration of effects), which models all justifiable measurement and analytic specifications in a single sample, to unpack the impact of researcher decisions when modeling these relationships. METHODS: Multiverse analyses were done with 3,648 unique models using a longitudinal sample of 6,639 Australian adolescents (aged ∼14.7-15.7, 2021-2022). RESULTS: Consistent co-occurrence of smoking or vaping and emotional problems was seen across unadjusted or only demographic-adjusted cross-sectional models (100 %). However, the temporality of relationships, choice of confounders, and operationalization of emotional problems substantially impacted findings. Emotional problems appeared to lead to reports of past 6 month smoking more-so than the reverse (88 % vs. 9 % unadjusted/demographic-adjusted), depression-focused measures yielded more consistent associations with smoking or vaping than anxiety-focused, and certain confounders (i.e., conduct, ADHD, other substances) explained most of the associations between adolescent smoking or vaping and emotional problems. Decision related to missingness or binary versus continuous outcomes did not meaningfully impact findings. CONCLUSIONS: While adolescent smoking or vaping and emotional problems commonly co-occur, methodological choices regarding timing, definitions, and confounding significantly influence the perceived strength of these relationships. Hence, such nuances demand careful consideration when interpreting evidence for policy.
