Individuals with substance dependence and MDD typically have more severe psychopathology as MDD is associated with prolonged course of dependence , greater treatment costs , and greater risk of suicide . The high prevalence of co-occurring MDD is also reflected in substance dependence treatment settings, with estimates ranging from 26-67% . Although many adults with co-occurring MDD do benefit from alcohol/drug treatment, they typically have worse outcomes than those without MDD . Despite these high prevalence rates and strong evidence of poorer treatment outcomes, few research studies have examined processes that predict post-treatment substance use within this population. Studies examining mechanisms of substance use treatment outcomes have not typically selected individuals with co-occurring psychiatric conditions, but have identified variables that likely translate to those with SUDs and MDD. As recognized by recent reviews, self-efficacy is a key construct involved in maintenance of abstinence following treatment for substance dependence , and is one of the most potent predictors of substance use treatment outcomes . Typically defined as the confidence/belief that one can resist using substances in high- risk situations, abstinence self-efficacy has consistently predicted treatment outcomes for both alcohol and drug users , in both inpatient/residential and outpatient settings . Studies also suggest enhanced self-efficacy is a common therapeutic element across treatment models as diverse as Alcoholics Anonymous , cognitive-behavioral therapy , and motivational enhancement therapy . Overall,marijuana grow system these studies indicate self-efficacy has a critical role in long-term treatment outcomes in substance-dependent population.

The role of social network characteristics in managing SUDs following treatment has also received considerable attention. As reducing contacts with users and developing sober relationships is frequently a component of treatment, of particular interest is whether the level of use in one’s social network actually predict post-treatment drinking and drug use. This has largely been supported in studies of alcohol users, as the overall frequency of drinking in the network, the proportion of network members abstinent, and the proportion of heavy drinkers in the network have all predicted post-treatment drinking . Network effects on drug use have not been studied in the context of formal treatment, but observational studies of drug users have had similar findings . Overall, these studies suggest social network characteristics could be a critical determinant of post-treatment use for patients with substance dependence and MDD. Despite extensive study of self-efficacy and social networks in general treatment samples, it is relatively unknown whether these process variables hold the same benefit for patients with SUDs and MDD, or if these patients can sustain self-efficacy and supportive social networks following formal treatment. However, some previous studies generally support the value of self-efficacy and social support among patients with SUDs and co-occurring mental health disorders. Among residential treatment patients with substance dependence and mixed co-occurring disorders including MDD, baseline self efficacy and general social support predicted six month substance use , and at follow-up self-efficacy was significantly associated with alcohol use . Among participants of dual-focus self-help groups, social support for abstinence mediated relations between greater group attendance and future substance use , while greater self-efficacy predicted greater quality of life . While these findings are promising, no known study has examined longitudinal, prospective effects of self efficacy or social networks on substance use following outpatient, professional psychotherapy for substance dependence and MDD.

Other treatment process variables were found to be less potent or more difficult to sustain for patients with MDD , which highlights the need to systematically evaluate effects of self-efficacy and social networks in patients with substance dependence and MDD.Within our sample, a potential moderator is living in environments that explicitly constrain alcohol/drug use, either via restricted access to substances or strong contingencies against using . Studies demonstrate that placement into halfway houses does achieve the intended effects of reducing substance use or achieving treatment milestones . Because substance use is more constrained by these environments, intrapersonal and interpersonal mechanisms like self-efficacy and social networks may be less crucial for individuals living in these “controlled contexts”. Investigation of these contextual effects is especially relevant for patients with substance dependence and MDD, who are especially likely to utilize residential or inpatient services . The moderating role of context has scarcely been investigated, but in one prior study the effects of the client-provider relationship on substance use were weaker for those in residential treatment . Thus, it is possible that controlled contexts moderate self-efficacy and social network effects, but to our knowledge no previous studies have investigated these relationships. The overall goal of this study was to examine prospective effects of self-efficacy and social network variables, the effects of controlled context, and their interactions in the prediction of post treatment substance use for veterans with substance dependence and MDD. Study participants received six months of either group Twelve-Step Facilitation or Integrated Cognitive-Behavioral Therapy for treatment of alcohol or drug dependence and MDD. Previous reports of this sample found that both groups increased in substance use during the one-year follow-up, with the ICBT group having comparatively less increase .

This study builds upon those findings by examining group differences and change over time in self-efficacy and substance-specific social network variables over the 12-months following treatment. We hypothesized that better post-treatment substance use outcomes would be predicted by greater self-efficacy and lower network substance use. Furthermore, we hypothesized that greater time in controlled context would be associated with lower alcohol/drug use, and that controlled context would moderate the effects of self-efficacy and network substance use on future drinking and drug use. This study involved secondary data analysis of veterans participating in a trial of outpatient group psychotherapy for co-occurring substance dependence and MDD, conducted at the San Diego Veterans Affairs Healthcare System . Participants met DSM-IV criteria for lifetime dependence on alcohol , cannabis , or stimulants with recent use, and DSM-IV criteria for MDD with at least one depressive episode occurring during a 3- month period of abstinence from substances. Exclusion criteria included dependence on opiates with intravenous administration, bipolar or psychotic disorder, living excessively far from the SDVAHS, or severe memory impairments limiting recall in assessments. With our current focus on the post-treatment period, we included all veterans completing at least one assessment from end-of-treatment to the one-year follow-up . The University of California, San Diego and VASDHS Institutional Review Boards approved the procedures for this study, which were explained in greater detail previously and are described briefly here. Study staff received referrals from the VASDHS dual diagnosis clinic, contacted veterans to conduct brief screenings, and met with eligible veterans to explain the procedures and obtain informed consent. Study participants consented to 6 months of group psychotherapy and 12 months of quarterly follow-up assessments, recording of group sessions,cannabis vertical farming psychotropic medication management appointments, random toxicology screens, and review of electronic medical records. All participating veterans consented to receive no additional formal treatment for substance use or depression for the duration of the treatment phase except for the psychotherapy and pharmacotherapy of the study, while participation in other formal interventions was allowed during follow-up.Veterans entered into group psychotherapy on a rolling basis, with start dates occurring every 2 weeks. After completing the intake assessment, veterans were sequentially allocated to the treatment condition with the next start date. Both interventions were manualized and were 6 months in duration, with twice-weekly sessions for 3 months followed by weekly sessions for the final 3 months. Group sessions were co-delivered by a senior clinician and doctoral trainee who were trained to criterion via manual review, direct observation, and weekly review and supervision. The protocol for Twelve-Step Facilitation was modified from TSF in Project MATCH for group delivery and targeting both drugs and alcohol. The Integrated Cognitive-Behavioral Therapy intervention was developed by adapting material from two empirically supported treatments: cognitive-behavioral relapse prevention from Project MATCH and group cognitive-behavioral therapy for depression . Both treatments were identically structured with a series of three modules, with each module covering a broad content area specific to 12-step or cognitive-behavioral principles. Each topic was introduced over a one-month block for the first 3 months of treatment, with review occurring in the final 3 months. Mean session attendance was not significantly different between TSF and ICBT . All veterans were offered pharmacotherapy management with VA physicians. Nearly all utilized medication management, with a mean attendance of 4.61 visits during treatment and 4.79 visits during follow-up. We utilized hierarchical linear modeling to examine process variables, controlled context, and their interaction in the prediction of post-treatment alcohol and drug use. The use of HLM was preferred for this study due to inclusion of multiple time points nested within individuals, use of both static and time-varying covariates/predictors, and inclusion of all available data via maximum likelihood estimation.

Maximum likelihood estimation is preferred when data contain missing values assumed missing-at random , and no significant differences were found any study variables between those with complete data and those with any missing data, supporting this assumption. With the exception of group and time effects, all predictor variables were grand-mean centered prior to inclusion in analyses. All statistical analyses were performed in Stata 10.1 .In preliminary models we examined static predictors of PDD and PDDRG to control for potential confounding variables , and then tested effects of treatment group, time, and the group x time interaction to model group differences in these outcomes. To examine group differences and time effects for self-efficacy, social network, and controlled context, these variables were analyzed as outcomes in HLM with group, time, and group x time as predictors. To examine prospective relations between process variables and substance use, lagged self-efficacy and each social network variable were used as time-varying predictors of PDD and PDDRG in HLM. Each social network variable was substance-specific . For these longitudinal analyses with time-varying covariates, the repeated measures contain confounded information about between-person and within-person differences, and disaggregation of these effects has been recommended when dictated by substantive interest . In this study, both effects were of interest: between-person differences or within person differences could predict future substance use. To disaggregate the between- and within-person effect, raw scores were de-composed into two variables representing the person-mean and time specific deviations from the person mean , and both variables were included in HLM to examine independence of these effects. To examine effects of controlled context, the controlled context variable was entered as a time-varying covariate of current PDD and PDDRG. The final HLMs examined interactions of lagged process variables with concurrent controlled context, to examine whether controlled context moderated effects of lagged process variables on PDD and PDDRG. Prior to testing core hypotheses, preliminary HLMs examined intake covariates of post-treatment PDD and PDDRG. Greater years of education predicted lower PDD , being employed predicted lower PDDRG , and pretreatment frequency of use predicted PDD and PDDRG . Models examining effects of treatment group, time, and the group x time interaction revealed that PDD and PDDRG increased significantly following treatment, independent of treatment group. Furthermore, the group x time interaction was statistically significant for PDD, replicating the previous findings of relatively greater 3 increase in substance use in the TSF group during follow-up . All subsequent models of PDD and PDDRG accounted for these significant covariate effects. By testing effects of treatment group, time, and the group x time interaction in HLM, the next series of models examined group differences in mean levels, change over time, and group differences in change over time in self-efficacy, social network variables, and controlled context during the 12-month post-treatment follow-up. As shown in Table 9, estimates of variance components revealed significant variability at both the between individual and within-individual level for self-efficacy and social network variables, justifying the use of HLM and disaggregation of these effects. A significant group x time interaction was found for social support for drug use , percent-network abstinent from drugs , and percent network using drugs . Inspection of means revealed that at month 6 the TSF group had relatively greater percent-network abstinent from drugs, with lower levels of mean network drug use and percent-network using drugs.