From January through May 2005 The Walsh Group, with the assistance of The Center for Survey Research (CSR) at the University of Virginia, conducted a national survey of people in recovery for substance abuse problems who were currently attending mutual-aid support groups. Respondents were contacted through mutual-aid support groups and data was collected using both mailed and on-line anonymous surveys. Dr. Atkins developed the questionnaire, pre-tested it with members of all the targeted groups, and revised it with the assistance of CSR staff. The mailed survey questionnaires and return envelopes included no identifying marks. They were distributed at group meetings and completed mailed questionnaires were returned to CSR for data entry. The web version of the survey was designed and maintained by CSR. To assure anonymity, respondents obtained an id and randomly generated password from one site and then proceeded to a second, independent site that housed the online survey site where they would use their id and password to enter the survey. The id and password enabled respondents to leave the survey and return to complete it later if need be. CSR conducted reliability checks for data entry and tracking, merged the web and mail surveys with an indicator as to source, and prepared the final data file for analysis. The authors then conducted the data analysis.
Identical survey questionnaires were administered via mail and on-line survey procedures. The on-line version of the survey took an average of 30.46 minutes to complete. T-tests revealed there were no significant differences between those who completed the mail survey and those who completed the survey on-line with respect to participating in one's primary self-help group (t = 0.099; p = .921) or number of days one has been sober (t = 1.87; p = .062). These are the two central variables in our analysis.
2.1 Sample Selection
The targeted groups included both 12-step groups (AA and NA) and non-12-step groups (SOS, SMART, WFS). All five of these targeted groups were invited to participate in the project through their national offices. SOS and SMART facilitated recruitment by providing national mailing lists of active groups and posting links on their websites to the online version of the survey. Active groups from SOS and SMART were randomly selected by state and the group coordinators on the lists provided by these groups were mailed questionnaire packets as well as scripts and instructions for distributing the surveys to participants at their meetings. Some additional mailed surveys were distributed at selected SOS and SMART meetings per these groups' requests. The 12-step groups, because their “The Twelve Traditions” preclude organizational involvement in research, could not provide mailing lists, but invited researchers to distribute surveys at open meetings. Thus, an alternative recruitment strategy was required and 12-step respondents were recruited through personal contacts in 12-step recovery and at open meetings. Personal contacts were given packets of questionnaires and asked to distribute them at meetings and through their own 12-step contacts. The research team also attended open meetings, asking permission from members prior to the meetings and then distributing surveys outside the meeting rooms after the meetings had concluded. An individual from the recovery community also volunteered to help recruit 12-step respondents on-line, and after getting permission from sites' hosts, posted information on the survey and the on-line survey link on two 12-step chat groups. WFS was included in the survey via the on-line version, which was advertised in their monthly newsletter and on their website. Surveys were mailed to those WFS members who preferred not to use the on-line survey and contacted us requesting a paper version of the survey. Members from other recovery groups who inquired about the survey were also invited to participate on-line. All respondents were provided informed consent in compliance with all National Institutes of Health regulations [45 CFR 46].
In total, 1,067 persons responded to the survey. Of these, 51 were not in recovery, 38 failed to indicate if they were in recovery, and another 54 did not identify a primary recovery group. These 143 cases were not included in the analysis. The remaining 924 cases are all people in recovery who identified a primary recovery group with which they were currently associated. Of the 924 usable cases, 270 completed paper surveys returned by mail (29.2%) and 654 completed the survey on-line (70.8%).
Respondents were asked what group they considered to be their primary group. Because of their similarities with respect to spirituality and religiosity, we combine AA, NA and other 12-step groups into a single “12-Step groups” category. Respondents were classified as members of either 12-Step groups (n=161), SOS (n=104), SMART Recovery (n=321) or Women for Sobriety (n=236). An additional 102 cases were classified as members of “other groups.” These included Moderation Management, Rational Recovery (RR no longer officially endorses the support group approach), LifeRing Secular Recovery, and others. These other recovery groups were too diverse in their approach to combine for analysis purposes. We therefore exclude these cases from our analysis here. Consequently, the sample used in our analysis included 822 respondents.
2.2 Sample Characteristics
The mean age for respondents was 47.0 years, ranging from 18 to 82. The sample was 58.8% female and 41.2% male. The gender distribution was skewed by the participation of Women for Sobriety in the survey, which is a women-only group. If we remove WFS from the sample, the breakdown is 58.1% male and 41.9% female. Our sample was primarily white (90.3%), with 9.7% being “non-white.” Fifty-one percent of the respondents were married (31.8% were in their first marriage), 20.7% were single, 3.4% were separated, 22.7% were divorced and 2.2% were widowed. The mean annual family income was approximately $55,000. Thus, the sample was primarily white and middle class.
Most of the respondents were long-time substance users before entering recovery, with 78.8% reporting having used their primary drug of choice for more than 10 years and another 8.4% having used for 7 to 10 years. Overall, 30.5% of the respondents reported having been in jail, 35.5% had been hospitalized for drug treatment, and 21.4% had been hospitalized for psychiatric treatment at least once in their lifetime. Nearly half (48.7%) had been in outpatient treatment at least once in their lifetime, with 16.5% doing so more than once. Only 27.7% reported having never relapsed after first entering recovery, with 40.4% reporting relapsing more than 3 times.
The respondents' first drug of choice was typically alcohol. Over three-fourths of respondents listed alcohol as their “drug of choice.” Prescription drugs (9.6%), marijuana (6.2%) and cocaine (3.9%) followed alcohol as the most common primary drugs of choice. The second drug of choice for respondents was marijuana (25.0%), prescription drugs (21.5%), alcohol (15.6%), and cocaine (9.2%). Just over 70% reported being multi-drug users. The vast majority of respondents (82.7%) said there was a history of substance abuse in their family and 51.8% said there was someone in their family with a current substance abuse problem.
While our sample was not selected randomly, it appears to be reasonably representative of the recovery-group population when compared to the few other studies of mutual-aid support groups available that had random samples. For example, in AA's 2004 random sample of its members they found that 89.1% were white and the average age of AA members to be 48 years (AA, 2006
). Similarly, Kaskutas' (1994)
survey of entire membership of Women for Sobriety found the average member to be white and 46 years of age. Connors and Dermen's (1996
; Christopher, 1992)
random sample survey of SOS members found them to be 99% white with an average age of 47 years. No previous surveys of SMART Recovery participants have been conducted. Hart's (2001)
national random sample survey of the recovery community found it to be 61% male and 39% female, middle aged, and with 79% being white; however, only 52% of those in Hart's survey said they participated in “self-help” groups. It appears that participation in mutual-aid support groups is more likely for people who are white and middle class, which is reflected in our sample. Sample limitations are difficult to overcome given the importance of anonymity for these groups. Most importantly, the current sample does provide adequate variation in the variables of central interest (i.e. religiosity, program participation, and sobriety). Thus, we consider the data, while limited, the best available data to address the question at hand.
2.3 Operational Measures
The analysis includes a series of regression analyses to determine if (1) respondents who participate in a mutual-aid support groups are more likely to remain sober and (2) if those respondents whose religious beliefs match those typical of their primary recovery groups were more likely to participate in recovery groups. The first dependent variable, sobriety, was measured by asking respondents the month, day, and year they “got clean and sober.” This date was transformed to the number of days the respondent had been “clean and sober.” The dependent variable is the natural log of the number of days the respondent had been sober. The natural log was taken to correct for extreme skewness. Once the log transformation was performed, the variable was normally distributed.
The second dependent variable, program participation, was measured using a nine-item scale that was created by combining seven five-point Likert-scale items on group related behaviors and two variables that measured the respondents' frequency of attendance in recovery support groups. The seven Likert items included questions on how often respondents participated in group discussions at meetings, led meetings, helped set up or cleanup at meetings, read group literature outside of meetings, helped newcomers, talked with members outside of meeting settings, and did things with other members that were not directly recovery related [e.g. dinner, the movies]. Values for each item were coded as never = 0, rarely = 1, sometimes = 2, often = 3, and very often = 4. An item asking respondents the number of meetings they attended in the past month was also included. This item ranged from 0 to 7 (7 or more times in the past month). Finally, an item asking respondents about the number of different groups they attend was included. This item ranged from 0 to 6 (five or more groups). The nine items were combined into an additive index (alpha = .800).
The independent variables of central interest for the analyses are the respondent's degree of religiosity and belief in a higher power. Fifteen five-point Likert-scale items were analyzed using a principal component factor analysis. The items loaded on two dimensions, and the two-factor solution explained 64.5% of the variance in the fifteen items. The first dimension represented the respondents' religiosity as reflected in items such as “Religion gives meaning to my life” and “I am sometimes very conscious of the presence of God.” The second dimension represented the respondents' belief in a Higher Power. The items reflecting this dimension included “my Higher Power helps me stay sober and drug free” and “my recovery group and its members are guided by a Higher Power.” The results of the factor analysis and the wording for each of the fifteen items are presented in .
Factor Analysis of Religiosity Items
Another variable of central theoretic interest is in which mutual-aid recovery support group the respondent participated. Respondents were asked which recovery group they considered their primary group. Those respondents who indicated their primary groups were SOS, SMART Recovery, Women For Sobriety or traditional 12-step programs (e.g., A.A. or N.A.) were included in the analysis. Mutual-aid support group was modeled as a series of dummy variables with traditional 12-step groups being the comparison group (i.e. SOS, SMART, WFS are compared to 12-Step).
In addition to these central variables, we control for age, number of three closest friends in recovery programs (0 = “none” to 3 = “all three”) and if the respondent was in a committed relationship (0 = not with someone, 1 = together with someone). We also control for the respondent's perception of the seriousness of their drug / alcohol problem. This item ranged from 1 (“not really a problem”) to 7 (“an extremely bad problem”). Gender was also initially controlled for in all of the models. Gender was not a significant predictor in any of the models; yet, since WFS is a women-only group, the inclusion of gender raised issues of multicollinearity. Since gender failed to predict any of our dependent variables (and was not related to them in a bivariate analysis), gender was removed from all of the models. Finally, we fit the interactions between religiosity and the respondent's primary recovery support group. The data were centered prior to creating the interactions to avoid problems of multicollinearity (see Aiken and West 1991
The analysis proceeds in three stages. Since the data are technically clustered (i.e. respondents are nested within primary self-help groups), in the first stage we determine if hierarchical models are needed by fitting a variance component model with group as a random effect. We then turn our attention to predicting the ultimate goal of any recovery program: namely, we predict the number of days respondents have been clean and sober. Finally, in the third stage, we investigate the factors that can predict the extent to which respondents participate in their recovery support groups.