Latent variable modeling of longitudinal and multilevel alcohol use data

This article demonstrates use of a latent variable model for longitudinal data which encompasses nested structures. Using Multilevel Latent Growth Modeling (LGM), levels of alcohol use and development of alcohol use over 4 years were examined among individuals (adolescents and parents) nested within families.

An LGM model was tested for alcohol use with a sample of 435 families (435 target adolescents, 203 sibling and 566 parents [168 fathers and 398 mothers]). Adolescents (targets and siblings) comprised 312 boys and 326 girls, with a mean (+/-SD) age at Time 1 (T1) of 13.69 +/- 1.95 years. It was hypothesized that there would be homogeneity in level and development of alcohol use among family members and heterogeneity in alcohol use and development across families. The effects of family status (single-parent, two-parent intact and stepparent families) and socioeconomic status (SES) on family levels of alcohol use were also examined.

Results suggested that stepparent families, and less educated and more economically disadvantaged families, had higher family levels of alcohol use and developed in their use of alcohol at a faster rate.

Findings suggest that the alcohol use of individuals in the same family is more alike than that of individuals from different families and that family alcohol use may be influenced by family-level variables such as family status or SES. Methods such as those presented, which incorporate family-level clustering, are likely to provide additional information regarding the etiology of alcohol use and risk factors for alcohol use within and across families.