Identifying risk profile for adolescent e-cigarette use: A sex-stratified machine learning analysis

Area of Science:

• Adolescent Health
• Public Health
• Machine Learning in Health

Background:

• Recent trends show higher e-cigarette use among adolescent females compared to males.
• Sex differences in adolescent e-cigarette use are documented, but underlying risk profiles remain understudied.
• Machine learning (ML) offers a novel approach to identify complex predictors of substance use.

Purpose of the Study:

• To identify and compare predictors of past 30-day e-cigarette use in adolescent females and males using an ML approach.
• To differentiate risk profiles for e-cigarette use based on sex in a sample of high school students.
• To inform targeted prevention and intervention strategies for adolescent e-cigarette use.

Main Methods:

• Analysis of cross-sectional survey data from 1829 9th graders (mean age 14.6 years; 54.7% female).
• Application of Gradient Boosting Machine (ML algorithm) to predict past 30-day e-cigarette use separately for females and males.
• Inclusion of 68 self-reported variables and identification of top 10 predictors per model using scaled importance scores.

Main Results:

• Overall past 30-day e-cigarette use was 3.6% (3.7% females, 3.5% males).
• For females, depression and post-traumatic stress disorder were leading predictors; for males, beliefs about and susceptibility to e-cigarette and cannabis use were key.
• Combined model predictors included cannabis use and peer e-cigarette use. Model performance was moderate.

Conclusions:

• Sex-specific risk profiles for adolescent e-cigarette use exist, highlighting the need for tailored approaches.
• Mental health factors are significant predictors for females, while substance use susceptibility and peer influence are critical for males.
• Findings emphasize the importance of sex-differentiated strategies in adolescent e-cigarette prevention and intervention programs.