Machine Learning of Smoking Relapse: the Role of Racial Differences and E-Cigarette Vaping Characteristics on Former...

Area of Science:

• Public Health
• Addiction Science
• Machine Learning in Health

Background:

• Machine learning models can identify complex factors in tobacco use transitions.
• A random forest model was developed to predict smoking relapse, considering racial disparities and vaping behaviors.

Purpose of the Study:

• To investigate racial differences in predictors of smoking relapse.
• To examine the role of vaping characteristics in smoking cessation outcomes across different racial groups.

Main Methods:

• Utilized data from the Population Assessment of Tobacco and Health (PATH) study.
• A random forest model was employed to predict smoking relapse among former smokers (n=4693) from baseline (Wave 5) to follow-up (Wave 6).
• Included 100 predictors encompassing demographics, psychosocial factors, health status, and substance use behaviors.

Main Results:

• Overall, 4.4% of former smokers relapsed within 4 years.
• High prediction accuracies (AUCs) were achieved across racial groups: Whites (0.77), Blacks (0.88), and Hispanics (0.70).
• Quit history, tobacco addiction, and marijuana use showed varying importance as predictors across racial groups. Vaping status and characteristics also influenced relapse differently by race.

Conclusions:

• Significant racial differences exist in factors predicting smoking relapse.
• Vaping characteristics play distinct roles in relapse prediction among different racial groups.
• Personalized interventions addressing unique social, behavioral, and health factors are essential for improving smoking cessation.