Surviving Drug Addiction

The Effect of Treatment and Abstinence on Mortality

Christy K. Scott, PhD; Michael L. Dennis, PhD; Alexandre Laudet, PhD; Rodney R. Funk, BS; Ronald S. Simeone, PhD


Am J Public Health. 2011;101(4):737-744. 

In This Article


The data we used were from a study that included a stratified sample of 1326 adults (participation rate: 85%) recruited between 1996 and 1998 from sequential patients admitted to a network of 22 addiction treatment programs operating on the west side of Chicago (10 outpatient drug-free programs, 5 intensive outpatient drug-free programs, 3 methadone maintenance clinics, 2 short-term inpatient clinics, 1 long-term inpatient program, and 1 halfway house).[28,35,44–47] Participants were reinterviewed at 6 months, 2 years, and annually thereafter for 9 years (follow-up rates per wave were 92%–96%). The follow-up was conducted from 1996 to 2007.

Baseline data on all 1326 participants and 131deaths (9.9%) over the course of the 9-year follow-up were available to address our first research question (centering on mortality rates overall and by subgroup). To address our second research question (centering on multivariate relationships between baseline and intervening variables), we focused on a subset of 1222 (92%) participants for whom we had all data from baseline, from the 6-month follow-up, and from at least 1 follow-up from the year before death or the final study observation (i.e., so that we could estimate the risk of mortality in the subsequent 12 months).


Baseline data showed that participants were mostly women (59.2%), aged between 30 and 49 years (68.7%; mean = 34.3 years), African American (87.5%), and unemployed (92.0%); most had never been married (65.3%). About half had earned a high school degree or its equivalent (48.7%), and 32.2% considered themselves homeless, with 11.7% having lived on the street or in a homeless shelter in the 6 months prior to intake. Sizable percentages of the participants reported histories of physical (24.3%), emotional (34.6%), and sexual (18.8%) victimization.

On average, participants had initiated drug use at age 16.8 years, had used regularly for 14.4 years, and had used regularly for18 of the preceding 30 days. Most (53.2%) reported prior addiction treatment, including 27.4% with 2 or more prior treatment episodes. More than 98% of participants had used multiple substances in their lifetime, and 76% had done so in the 30 days before intake. The most common substances used 5 or more times in the 30 days before intake were cocaine (34.8%), alcohol (20.7%), heroin (31.1%), and marijuana (8.5%). Few participants (3.8%) reported injecting drugs in the 6 months prior to intake. Other problems included major depression (36.4%); generalized anxiety disorder (36.3%); a history of criminal justice involvement, including arrests (76.9%), convictions (49.9%), and incarceration of 1 or more months (39.1%); and current probation or parole status (25.0%).

Data Collection

The main study instrument was an augmented version of the Addiction Severity Index,[48,49] which includes questions on age, lifetime and past-month problem severity, employment, family situation, and psychiatric functioning. We modified the instrument to collect more detailed data on treatment and incarceration histories, service use, high-risk behaviors, mental distress, pregnancy, illegal activity, criminal justice involvement, recovery environment, drug use practices (or lack thereof) of friends, impact of substance use on relationships, type of disability or chronic condition, body mass index, motivation, and coping.

The Addiction Severity Index drug (a=0.71), alcohol (a=0.86), and psychiatric (a=0.83) composites have been shown to have good internal consistency,[28,47] with selfreported use of outpatient, methadone, and residential treatment in the first 6 months after intake correlating well (>0.7) with treatment records. In a substudy of 259 participants, there was good concordance between urine tests and self-reported rates of substance use both overall (73%) and specifically with respect to opioids (82%), cocaine (68%), and marijuana (68%). The test–retest reliability was also good (j values ranged from 0.62–0.86) across the subscales, meeting or exceeding the instrument's published rates of internal consistency (0.6–0.9) and testretest reliability (0.5–0.9).[49] The Addiction Severity Index composite scores and other measures have been shown to be sensitive to duration of abstinence as well.[35]

The follow-up protocol[50] involved routine contact with each participant between interviews. If the participant could not be contacted, we attempted to reach collaterals and service agencies with which the participant had dealings to reestablish contact. If we learned from these sources or public records that a client had died, we recorded the period of death (i.e., at what point in the study the death occurred); however, we did not record the date of death or information on the cause of death. Participants received $50 for completing for the interview, $10 for completing it on time, $15 for a urine sample, and $5 for confirming their appointment.

Analytic Approach

For our first research question (focusing on mortality rates overall and by subgroup), we limited the analysis to baseline predictors measured at study intake. For our second research question (focusing on multivariate relationships between baseline and intervening variables), we included variables from baseline, the first 6 months of treatment, and the remaining months of the study (months 7–96) so that we could distinguish between early and later treatment. These variables were used to predict the likelihood of achieving at least 1 year of abstinence. Only variables that remained significant in the multivariate model were included. This analysis was then repeated with the addition of variables representing percentage of time abstinent and years of continuous abstinence to predict the risk of mortality in the subsequent 12 months. SPSS version 17 (SPSS Inc, Chicago, IL) was used to conduct all logistic regression analyses.

Odds ratios (ORs) significantly greater than 1 suggest an increased likelihood of the respective dependent variable, and odds ratios significantly less than 1 suggest a reduced likelihood of the respective dependent variable. To test the extent to which years of sustained abstinence mediated the outcomes assessed, we followed the recommendations of Baron and Kenny[51] and tested the final model for predicting mortality with and without this variable to evaluate the impact on the overall fit and individual parameters.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.