The Association between Non-Partner Violent Behavior and Substance Use over Time: Findings from an Accelerated Cohort of Adolescents in Southeast Michigan

Introduction: Despite an overall decline in rates of adolescent alcohol and marijuana usage since the 1990s, considerable variability exists in individual usage patterns over time. Violent victimization and violent behavior are each associated with AOD use, but these associations are often examined cross-sectionally or retrospectively. Accelerated longitudinal designs can capture a wider developmental span than traditional cohort designs over the same period. We describe the non-linear trajectories of alcohol and other drug (AOD) use in an accelerated cohort of adolescents and examine violence correlates of both AOD intercepts and slopes.

Methods: Data were drawn from the Strengthening Healthy Adolescent Relationships and Environments (SHARE) study, an accelerated longitudinal design comprised of two cohorts (6^th grade, N = 585, Female = 51.6% (302), White = 55.9% (327); 9^th grade, N = 647, Female = 52.6% (340), White = 62.0% (401)) followed over four years. The dependent variable is a mean of four substance use indicators (past year alcohol, marijuana, illegal drug use and binge drinking) assessed annually over the four years. Independent variables included non-partner violent behavior, adverse childhood experiences (ACEs) score, peer support, parental supervision, depression/anxiety, school/neighborhood safety, sex, and race. Using linear mixed-effects models, we first examined three forms of change in AOD use over four waves of data collection: 1) an intercept only model, 2) a linear change model and 3) a non-linear (i.e., quadratic) change model. After determining the functional form of AOD use patterns, we included covariates associated with both baseline AOD use (i.e., intercepts) and changes over time (i.e., time by covariate interactions).

Results: A quadratic slope provided the best fit to the unconditional AOD growth model. Overall, AOD use increased over time with a small acceleration (b_Grade^2=0.016, SE= 0.002, p < .001) over the developmental period. Significant variability in both intercepts (t₀₀=.006) and slopes (t₁₁=.023; t₂₂=.001) confirmed the ability to model individual variability in change in AOD use rates over time. Predictors of baseline substance use included non-partner violent behavior, parent supervision, depression/anxiety, and ACEs. When examining substance use trajectories, only adolescent perceptions of neighborhood and school safety were associated with change over time.

Conclusions: A hisotry of violence experiences and behavior were predictors of baseline substance use, but were less informative when considering changes in use from grades 6-12. The two variables associated with change in substance use over time were contextual influences, rather than individual level factors. These findings support the need for multi-level interventions that consider the multifaceted nature of violence and safety perceptions. We will discuss the relative strengths and weaknesses of similar designs for the study of adolescent risk behavior.