A within and Between Person Latent Variable Analysis Between Substance Use, Criminal Behavior, & Environmental Risk Among Emerging Adults in Treatment

Literature that examines the longitudinal associations between crime and substance use has found significant crossed-lagged relations, such that crime predicts substance use, and substance use predicts crime. However, no study has included the effects of environmental risk (ER; e.g. delinquent peers, high risk living situation, and high risk work environment) on criminal behavior (CB) and substance use (SU). Furthermore, these studies use auto-regressive crossed-lagged models that do not partition within- and between-person variance, essentially treating the effects as identical across the two levels of analysis. This study aims to address these shortcomings by considering the longitudinal effects of ER on both CB and SU while using auto regressive latent growth trajectory with structured residuals (ALT-SR) to partition within- and between-person variance. A sample of 3,429 emerging adults entering substance use treatment to model longitudinal associations between SU, CB, and ER over a one year period measured quarterly. Participants were, on average, 20 (SD = 2.3) years old, primarily male (70%), and 37% White. Participants reported, on average, 11 days of alcohol use in the past 90 days, and 31 days of probation. Our growth parameter included a piecewise (or jump) parameter from baseline to 3 months and a piecewise growth (linear growth after jump) centered at the 3 month follow up. We allowed baseline to 3 month auto-regressive paths to be freely estimated due to steep declines on all constructs. All within person cross lags were constrained. Between person effects were estimated by allowing latent intercept and growth parameters to correlate. We used reductions in -2 Log likelihood, to asses model fit. Our final model had excellent model fit: CFI = .017, RMSEA = .981. Results indicated a between-person relationship exists between CB, ER, and SU (and all other combinations). However, the within-person cross-lagged associations indicate no significant relationship exist between SU and CB (B = .003) (and vice versa (B = .130)). Interestingly, it appears that ER and SU have a cross lagged relationship (B = .077, p < .01; B = .052, p < .01) with ER also associated with increased CB (B = .010, p < .01). This study finds that criminal behavior may not be a substantial predictor of substance use, within-person, whereas a person’s environment may be more important. Clinicians working with EAs may wish to specifically target environmental risk such as time spent with deviant peers or use of group (or family) therapy to aid in high risk living situations.