Our study of social network effects confirms that adults with SUD and MDD will likely benefit from efforts to restructure their social networks by increasing the number of abstinent/non-using supports and decreasing the number of contacts who use substances regularly. Results also suggested that both regular maintenance of positive social networks and avoiding momentary shifts that increase the density of regular users are likely important for avoiding future increases in drinking. For individuals who are unable to enact sustained changes in social networks, our research suggests that placement into constrained environments can assist in attenuating the negative influence of a social network comprised of greater numbers of substance abusers. While this series of studies makes important research and clinical contributions, especially with respect to the treatment and long-term recovery of patients with substance dependence and MDD, overall limitations should be noted. Perhaps most notable among these limitations are the restricted demographic characteristics of our sample which curtails the immediate generalizability of these findings. This is a common limitation of clinical trials of treatments for addictive disorders , and while these results may be widely applicable due to the high prevalence of MDD in substance use treatment settings , replication in other samples is needed before generalizing these findings to a population with a wider range of demographic characteristics. Many of our patients also had post-traumatic stress disorder. While we tested effects of PTSD on substance use outcomes, we did not explore more intricate effects that are of interest due to the high prevalence of PTSD in this population. Temporal precedence between variables was not always tested , and even models demonstrating temporal precedence did not fulfill all criteria for examining mechanisms of change . There is still some debate as to whether some of the process variables we studied are true mechanisms,indoor cannabis grow system or just risk markers of some underlying characteristics that better enable patients to limit drinking and drug use .
Regardless of these distinctions, our studies make important contributions to the literature because these variables have not been studied extensively in patients with SUDs and co-occurring psychiatric disorders, but further research is needed to demonstrate these processes can be manipulated in ways that improve the effectiveness of interventions. Because these studies involved secondary analyses of existing data, we were somewhat limited by characteristics of measures utilized in the original clinical trial. Our measures of 12-step attendance and affiliation were brief and did not differentiate between different meetings , which may have allowed better investigation of disparate findings with respect to alcohol vs. drug use. Our social network measure did not capture features of social support examined in prior studies, such as whether network members actively supported patients’ abstinence efforts . Furthermore, although we examined complex mediating and moderating effects on long-term substance use, there are potential relationships among mediating variables that we did not investigate.Considering the results of these four studies in the context of prior research in substance-dependence samples, there are numerous interesting lines of research for future inquiry. While we found that impairment predicted overall levels of 12-step affiliation, we did not test whether impairment relates to specific 12-step behaviors or impacts rates of post-treatment changes in 12-step attendance or affiliation. Because the rate of change in 12-step affiliation may be especially important, as demonstrated in Study 2, future research might further investigate the role of impairment on changes in 12-step affiliation in patients with substance dependence and MDD. Prior studies also found that neurocognitive functioning moderated social network effects , and given that social networks predicted substance use and impairment moderated the effects of other contextual variables , this question is worth exploring in patients with substance dependence and co-occurring psychiatric disorders.
Studies have demonstrated that patients’ pre-existing social network characteristics may moderate the effects of certain interventions, such that treatments specializing in altering the structure of social networks are especially useful for patients with networks comprised primarily of substance users . It is currently unknown whether social networks or many other baseline characteristics impacted whether patients in our sample had better outcomes from TSF or ICBT, which may help guide decisions about the appropriateness of certain interventions for patients with substance dependence and psychiatric disorders. Further work is needed to determine if particular aspects of psychosocial therapy are responsible for initiating change in mediating or moderating variables, and how to optimally package interventions for dissemination. A prior study of alcohol-dependent patients found that receiving a specific craving module moderated relations between negative mood and drinking . Given that patients in our study received distinct modules that focused on different aspects of coping, and depression predicted future substance use, it may be worth investigating whether receipt of certain therapy modules had intended effects on reducing “self-medicating” patterns of drinking. Homework compliance has been found to predict outcomes in CBT for cocaine dependence and may be an important mechanism of intervention effects on substance use , suggesting that compliance with group therapy homework may have influenced therapeutic process variables in our sample. By identifying modifiable variables that predict future substance use , our research identifies important proximal targets for newly developed interventions or modifications to existing therapies. Since the publication of the short and long-term results of our trial, there is somewhat stronger evidence for the efficacy of integrated psychosocial therapies , but given the historically poor treatment outcomes for patients with substance dependence and MDD, further development and refinement of interventions is necessary to improve the long-term treatment outcomes within this population. The corresponding predicted reduction from the Poisson regression is approximately 2.1 percent for the minimum wage and 2.3 percent for the EITC. However, the precision of the estimates is too low to conclude that this difference is statistically significant. For additional robustness, we estimate augmented specifications controlling for state linear and quadratic time trends. The results are qualitatively consistent across these specifications ; however the precision of the estimates is reduced to the point where the estimated effects of the EITC are no longer statistically significant, possibly reflecting the limited variation in state policies during the sample period. Next, we present the estimated event study models of suicide deaths.
Figure 2 plots the estimated event time coefficients together with 95 percent confidence intervals. Panel presents results for the minimum wage. Recall that if the parallel trends assumption holds, we should expect the data to exhibit parallel pre-trends, i.e. the estimated event time coefficients should not be different from zero for the years leading up to a minimum wage increase . Overall, point estimates are indeed small in magnitude during the pre-period; and they are not significantly different from zero at the five percent level. At time 0, the estimated event time coefficients exhibit a significant discontinuous downward shift, consistent with the negative effect of the minimum wage presented in Tables 2-4. Separating the sample by gender, minimum wage event study estimates for men are somewhat more concerning – while the estimated pre-trends are not statistically significantly different from zero,PIPP horticulture the point estimates are nonetheless consistently positive. Such differential pre-trends could mean that the negative effect for men is biased downward, reflecting in part differential mortality patterns in states that implement higher minimum wages. For women, estimated pre-trends are small and close to zero, supporting parallel pre-trends. Moreover, the drop at time zero is statistically significant at the five percent level. Unlike the minimum wage, which raises pre-tax wages at the time of implementation, the direct impact of state EITC policies on disposable earnings may operate with a lag, as eligible families receive EITC payments only after filing taxes for the previous year. Absent any labor supply response, state EITCs would start affecting outcomes only in their second year, which is the first year eligible workers receive the additional payments. Meanwhile, the research consensus suggests that higher EITCs have positive employment effects, especially among single mothers. For these groups, expanded EITCs could have an additional, contemporaneous effect on pre-tax income as well as on associated downstream outcomes, as workers increase their labor supply knowing they will earn a larger EITC payment when they receive their tax refund the following year.The models find parallel pretrends for the pooled sample as well as for men and women separately. In the pooled sample, a small negative effect appears in year 0 , followed by a discontinuous downward shift in estimated event time coefficients the following year. This pattern is consistent with the effects of the EITC on suicides operating primarily through increased tax refunds in hand — as people start receiving larger tax refunds once the policy has been in effect a full year. For men, while there are no effects on suicides in year 0, event time coefficients drop sharply in year 1. For women meanwhile, the coefficient path starts falling immediately at year 0 followed by larger negative effects in year 1 and later years. This pattern is consistent with the literature that finds that positive labor supply responses to the EITC are found mainly among women. Appendix figures and show estimated results with further model varieties: Appendix figure shows results for minimum wages, but incorporating additional minimum wage changes by reporting outcomes for less than the full [-5,4] window around the policy change. Appendix figure presents results from the more parsimonious event study specification of equation . For the minimum wage, this specification indicates no significant shift in male suicide mortality around minimum wage changes. Meanwhile, estimated minimum wage event studies for the pooled sample and for women are remarkably consistent across the two specifications. For state EITCs, results are similar across specifications. To summarize, the estimated event study models indicate that the parallel pre-trend assumption holds, supporting our identifying assumption of parallel trends. In addition, the patterns indicate negative causal effects: the number of suicides tends to drop sharply after the implementation of higher minimum wages and state EITC. Our analysis to this point has focused on mortality outcomes of individuals with high school or less education, who have greater exposure to minimum wages relative to our placebo sample of individuals with a bachelor’s degree or higher. This same intuition should hold more generally: within the sample of less-educated adults, reductions in suicides should be larger among groups that are more exposed to the policies we study. To test this prediction, we use earnings and hours data from the CPS MORG to estimate exposures to the minimum wage for various groups of workers with high school education or less. We slice the sample by gender and age , yielding 10 sub-samples. We define group-level exposure to the minimum wage as the share of workers who earn less than 110 percent of the current minimum wage. To capture exposure to the EITC, we use the CPS ASEC, calculating for each demographic group the share of workers who receive the credit. We then estimate the panel models of suicide deaths from equation for each sub-sample. Intuitively, if minimum wages and EITCs reduce suicide deaths by raising incomes of affected workers, estimated effects should be larger and more negative for groups that have higher exposure. That is, the estimated effects should be negatively correlated with exposure. Conversely, a lack of correlation between effect size and exposure would provide evidence against our hypothesis that higher minimum wages reduce suicides by raising incomes of low wage workers. Figure 3 plots the estimated effects on suicide against exposure. The top panel shows effects for minimum wages, while the lower panel shows effects for EITCs. For both policies, the figure indicates that effect estimates and exposure are negatively correlated: on average, populations with higher exposure tend to experience more substantial drops in suicide. The line of best fit is downward sloping; for minimum wages, the slope is significantly different from zero at the 1 percent level, while the slope for EITC exposure is significant at the 5 percent level. We also find similar downward patterns when we plot effects versus exposure separately for men and women . To summarize, Figure 3 indicates that the reduction in suicides is greater among the groups that are more likely to be affected by higher minimum wages. This finding lends support to our hypothesized mechanism that minimum wages reduce suicides by lifting low-income groups out of poverty.