Multilevel Contextual Influences and Nonuse of Alcohol, Tobacco, and Other Drugs: A Latent Class Analysis of Substance-Free Youth

Author:
Yoder, Laura A. G., Nursing - Graduate School of Arts and Sciences, University of Virginia
Advisor:
Kulbok, Pamela, School of Nursing, University of Virginia
Abstract:

Currently very few studies have examined the co-existence of ecological health-negating and health-promoting influencing factors for nonuse, experimentation, and regular use of alcohol, tobacco, and other drugs (ATOD) among adolescents. Understanding which factors are most strongly related to hidden subgroups (latent classes) in the adolescent population, as well as which factors characterize the most homogeneous latent classes can point interventionists to tailoring variables and target subgroups, increasing the effectiveness of health promotion and substance use prevention programs.
This descriptive probabilistic study used archival data from the National Longitudinal Study of Adolescent Health (Add Health), waves 1-4, to identify latent classes of youth according to their patterns of ecological influencing factors. The latent class structure among nonusers, experimenters, and regular/risky ATOD users was tested. The probability of nonuse, moderate use, or regular/risky use during adulthood given latent class membership during adolescence was ascertained. Covariates included age, gender, race/ethnicity, family socioeconomic status (SES), residential location, and neighborhood SES. Latent Class Analysis (LCA) with covariates and LCA with a distal outcome were conducted using SAS.
For the overall sample (N=4,198), a 3-class model fit the data the best. Self-regulation, peer substance use, and school connectedness had the strongest associations with the latent variable. Age, family SES, and neighborhood SES were significant covariates. Since measurement invariance did not hold across subpopulations of nonusers, experimenters, and regular/risky ATOD users, the model was fit separately for each group. Each subpopulation fit a 2-class model the best, and the proportion of females within each class was significantly different when comparing the two classes of regular/risky users. Finally, relative to the overall population prevalence, regular/risky substance use during adulthood had the greatest probability (0.90) given membership in the “low self-regulators with mixed influences” class, and adult substance nonuse had the greatest probability (0.29) given membership in the “health-promoting influences” class.
The results suggest that interactions among health-negating and health-promoting factors across youths’ ecology do account for important subgroups in the population. The strongest latent class indicators, i.e., self-regulation, peer substance use, and school connectedness, could be useful for tailoring interventions, while the latent classes can guide interventionists towards target subpopulations. More research is needed to understand the transitions in latent class membership occurring over time, and to identify other strong predictors of substance nonuse extending through adolescence and into adulthood.

Degree:
PHD (Doctor of Philosophy)
Keywords:
adolescents, substance nonuse, latent class analysis, health promotion
Language:
English
Rights:
All rights reserved (no additional license for public reuse)
Issued Date:
2014/04/04