Methods: Structural equation modeling (SEM), was used to examine a parallel mediation model and the direct and indirect effects of trauma symptomatology, ASE, and SSR on post-treatment intake substance use outcomes after controlling for relevant covariates. The latent constructs of interest were measured using the following: trauma symptomatology was assessed using the 40-item TSC-40; ASE was gauged with the 20-item Drug and Alcohol Abstinence Self-Efficacy Scale; and SSR was captured with the nine-item Social Support for Recovery Scale is a nine item measure capturing the perceived experience of sobriety assistance in the form of emotional concern, and/or information women had from family, friends and treatment related associates. Composite mean scores for each constructs’ respective sub-scales were utilized in the measurement and SEM models. Since the outcome variable substance use 12 months’ post-treatment intake is binary, the SEM model used the weighted least squares estimator with mean and variance adjustments (WLSMV) was used, which computes ordinary least squares parameter estimates for continuous outcomes and probit parameter estimates for categorical outcomes. Model fit was evaluated with the robust WLSMV chi-square p > .05, comparative fit index (CFI) and Tucker-Lewis Index (TLI) > .90, root mean square error of approximation (RMSEA) < .08, and the weighted root mean square residual (WRMR) < .10. A bootstrap approach in which the standard error is computed based on 5000 bootstrap replicates was used to test for a significance of mediated effects.
Results: Data for this examination were collected from a NIDA funded longitudinal study. 62% of the women identified as Black; 88% of the women did not have more than a high school education; 74% were dependent on government subsidies; and 73% had a co-occurring mental health disorder. Results of the measurement model revealed excellent fit, χ2=112.33, df=84, p=.02; CFI=.96; TLI=.97; RMSEA=.03, WRMR=.64. All factor loadings were strong, .48-.94 (p< .01), and all latent constructs were correlated (p< .05). The full SEM model retained a parsimonious fit, χ2=228.2, df=190, p=.03; CFI=.96; TLI=.96; RMSEA=.02, WRMR=.80. Moreover, ASE was a significant partial mediator of the relationship between trauma symptomatology and post-treatment substance use (95% CI .002 - .096).
Implications: These results suggest that clinicians may want to utilize interventions aimed at increasing ASE, that may assist in the development of resiliency skills in early recovery, mitigating the influence of trauma symptomatology. Future research will want to expand on these findings, evaluating more robust measures and the influence of time on these interrelated constructs and their influence on relapse in early recovery.