We assessed mortality rates (overall and by subgroup) in the study sample as well as relationships between the different categories of variables assessed.
Baseline Predictors of Mortality
Of the 1326 participants, 131 (9.9%) died over the course of the 9-year study (11.0 per 1000 person-years). Table 1 presents mortality rates and bivariate odds of mortality over the course of the study as a function of sociodemographic characteristics and functioning at intake. Being older than 30 years nearly doubled the risk of mortality (OR=1.98), whereas being African American was associated with reduced odds of mortality (OR=0.52).
The risk of mortality increased with more years of alcohol use and more years of opioid use prior to baseline. We also found positive relationships between mortality and a longer amount of time from first use to first seeking treatment and having a history of overdose or delirium tremens. In addition, individuals with physical disabilities, chronic medical illnesses (e.g., seizures, asthma, emphysema, high blood pressure, heart disease, cirrhosis, pancreatitis, diabetes), hospitalizations in the 6 months before intake, and Addiction Severity Index medical composite scores above the median were at significantly increased risk of mortality, as were those who were living alone, were living below the poverty line (as defined by the US Department of Health and Human Services), had engaged in any illegal acts for money, or had been charged with a violent act in the 6 months before intake.
Predictors of Sustained Abstinence
Although percentage of time abstinent in the overall sample increased from 55% during the 6 months before intake to 79% at the end of the study period, only 418 of the 1222 participants (34%) included in the multivariate analysis achieved 1 or more years of abstinence. Odds ratios for the multivariate logistic regression tests associated with sustained abstinence or death in the subsequent 12 months are displayed in Table 2. The rows show the predictors that were significant in the multivariate model after stepwise selection. The predictors are organized by time frame: baseline, months 0 to 6 (initial treatment response), and months 7 to 96 (long-term response). At baseline all measures from Table 1 were considered, but only those that remained statistically significant in the multivariate model are included in Table 2. For the latter 2 time periods, we also considered number of substance abuse treatment episodes, percentage of time in treatment and in the hospital, and involvement in illegal activity for money.
Examination of the model including proximal measures revealed that no baseline risk factors were significantly related to the likelihood of achieving sustained abstinence. However, a higher number of substance abuse treatment episodes during the first 6 months of the study (OR=1.32 per episode) and a longer amount of time spent in treatment over the course of the study (OR=1.42) increased the likelihood of achieving sustained abstinence. By comparison, after the first 6 months of the study, the likelihood of sustaining abstinence decreased as the number of subsequent treatment episodes increased (OR=0.75 per episode), the percentage of time spent hospitalized increased (OR=0.14 per 10-percentage-point change), and the percentage of days involved in illegal activity for money increased (OR=0.77 per 10-percentage-point change).
Predicting Time to Mortality
Table 2 also provides odds ratios significantly associated with mortality in the subsequent 12 months. This analysis included all of the variables from the baseline model just discussed, percentage of time abstinent in the initial and subsequent periods (as measures of harm reduction), and years of sustained abstinence at the final observation before mortality or the final observation minus 1 year. Mortality in the subsequent 12 months was associated with older age at intake (OR=1.82), the presence of a preexisting chronic illness (OR=1.85), and the amount of time a person engaged in illegal activity for money in the 6 months prior to intake (OR=1.14).
Although none of the initial treatment response variables remained significant in the multivariate model, mortality in the subsequent 12 months was associated with percentage of time spent in the hospital (OR=14.45) and in substance abuse treatment (OR=1.68) over the long term (months 7–96). We interpreted these 2 findings as markers of nonresponse to treatment. Over the long term, the likelihood of mortality decreased with additional substance abuse treatment episodes (OR=0.68), more time abstinent (OR=0.74), and more years of continuous abstinence (OR=0.81). Thus, the extent to which people with substance abuse problems were readmitted combined with their response to treatment (as both a percentage of time and a continuous period) was associated with a reduced risk of mortality.
Relationships between Risk, Treatment, Abstinence, and Mortality
Figure 1 provides a graphic summary of the numeric findings presented in Table 2 to help illustrate the complex relationships described here. A complex relationship existed between treatment, abstinence, and mortality in the multivariate analyses. The likelihood of mortality decreased directly as the total number of treatment episodes increased (OR=0.68 per episode) but increased with increasing percentage of time in treatment (OR=1.68 per 10-percentage-point change). Treatment also had an indirect effect via the reduced risk of mortality associated with both continuous abstinence (OR=0.81 per year) and percentage of time abstinent, even when it was episodic (OR=0.74 per 10-percentage-point change). The likelihood of sustained abstinence increased with increases in the number of treatment episodes in the first 6 months (OR=1.32 per additional episode) and in the percentage of time in treatment over the study period (OR=1.42 per 10-percentage-point change). However, the number of times a person returned to treatment between 6 months and 8 years after intake was associated with a lower likelihood of eventually achieving sustained abstinence (OR=0.75 per additional episode). Thus, the timing of readmission matters.
Summary of the relationships observed: adults in addiction treatment programs, Chicago, IL, 1996–2007.
We used Baron and Kenny's approach to assess the extent to which years of sustained abstinence mediated the relationships between risk, treatment, and mortality. According to the first criterion, the hypothesized mediator (years of abstinence) and each of the other predictors had to be significantly related to the dependent variable (mortality). The second criterion required 1 or more of the predictors to have a significant relationship with the mediator. As shown in Table 2 and Figure 1, these 2 criteria were met for a subset of 3 variables: percentage of time hospitalized, number of substance abuse treatment episodes, and percentage of time in treatment (10-percentagepoint units).
The third criterion required a change in 1 or more of the relationships between these predictors and mortality when the model was estimated with and without the mediator (data not shown); it also required that there be a better overall fit with the mediator in the model. The odds ratios for predicting mortality in the subsequent 12 months changed significantly (from the model without the mediator to the model with the mediator in Figure 1) for each of these variables (from 18.4 to 14.45 for percentage of time hospitalized, from 0.71 to 0.68 for number of substance abuse treatment episodes in months 7 to 96, and from 1.63 to 1.68 for percentage of time in substance abuse treatment). Moreover, including years of sustained abstinence improved the overall model fit with respect to predicting mortality in the subsequent12 months (χ2 1=12.11, P<.01). Thus, years of sustained abstinence qualified as a significant mediator of mortality risk in the subsequent 12 months.
Am J Public Health. 2011;101(4):737-744. © 2011 American Public Health Association
Cite this: Surviving Drug Addiction - Medscape - Apr 01, 2011.