Grant: Catalyst Grant: Health Effects of Vaping

Problem: Understanding person-level drivers of current e-cigarette use (vaping) is crucial to guide tobacco policy, but prior studies have not fully identified these drivers due to the reliance on cross-sectional data, small sample sizes in many studies, lack of generalizability, and limitations of traditional data analyses. We address this gap in knowledge by using machine learning methods including methods capable of learning from multiple data sources. Methods: We will extend our development of machine learning approaches to identify predictors of vaping onset and dependence. Data will be included from multiple longitudinal cohort studies that assess vaping. Person-level characteristics will be explored as candidate predictors. The purpose of this study is (1)To identify and rank the importance of characteristics that are predictive of vaping onset and dependence using a machine learning model within multiple datasets. Using this model, we will also assess the probability of effects that vary across subgroups (age, sex/gender, race);(2) To assess how differences in effects between surveys (population, geographies, measures) impact the predictors of vaping onset and dependence; (3) To implement and further develop integrative data analysis procedures to allow for machine learning across multiple studies with increased power. Impact: Findings from this study will help inform current public health debates on e-cigarette use and guide policy and clinical interventions. This study will be an important endeavor and first of its kind in the application of machine learning and integrative data analysis to vaping, which will pave the way for future use of this analytical technique in this field.