Effects of Long-Term AA Attendance and Spirituality on the Course of Depressive Symptoms in Individuals with Alcohol Use Disorder

Abstract

Alcohol use disorder (AUD) is associated with depression. Although attendance at Alcoholics Anonymous (AA) meetings predicts reductions in drinking, results have been mixed about the salutary effects of AA on reducing depressive symptoms. In this single-group study, early AA affiliates (n=253) were recruited, consented, and assessed at baseline, 3, 6, 9, 12, 18 and 24 months. Lagged growth models were used to investigate the predictive effect of AA attendance on depression, controlling for concurrent drinking and treatment attendance. Depression was measured using the Beck Depression Inventory (BDI) and was administered at baseline 3, 6, 12, 18 and 24 months. Additional predictors of depression tested included spiritual gains, [Religious Background and Behavior questionnaire (RBB)] and completion of 12-step work [(Alcoholics Anonymous Inventory (AAI)]. Eighty-five percent of the original sample provided follow-up data at 24 months. Overall, depression decreased over the 24 month follow-up period. AA attendance predicted later reductions in depression (slope=−3.40, p= 0.01) even after controlling for concurrent drinking and formal treatment attendance. Finally, increased spiritual gains (RBB) also predicted later reductions in depression (slope=−0.10, p=0.02) after controlling for concurrent drinking, treatment, and AA attendance. In sum, reductions in alcohol consumption partially explained decreases in depression in this sample of early AA affiliates, and other factors such as AA attendance and increased spiritual practices also accounted for reductions in depression beyond that explained by drinking.

Introduction

There is now strong evidence that attending Alcoholics Anonymous (AA) predicts increased alcohol-abstinence at a later time for many, but not all, problem drinkers (e.g., Magura et al., 2013; Tonigan and Rice, 2010; Kelly et al., 2011; Emrick et al., 1993). Findings are mixed, however, about potential secondary benefits associated with AA affiliation, benefits that may include reduced health care utilization and costs (Humphreys et al., 2004), changes in tobacco use (Reich et al., 2011), increased purpose in life (Oakes, 2008), and improved quality of life (Laudet, Morgen, & White, 2006; Spalding and Metz, 1997). Given current cost-containment pressures in the United States, the documentation of secondary benefits (and the estimation of their relative magnitude) associated with community-based 12-step attendance has become increasingly important.

For a variety of reasons, the investigation of changes in clinical depression has been of special interest as a secondary outcome among adults attending AA. First, 12-step programs such as AA make a clear ideological distinction between abstinence and sobriety (AA World Services, 2014). The measurement of changes in clinical depression offers an intuitively appealing and face-valid construct to investigate the higher-order objective of sobriety. Stated differently, decreases in depression might be a marker for the increases in well-being associated with sobriety rather than abstinence alone. Second, many theorists regard the experience of negative affect to be a primary precipitant for alcohol relapse (e.g., Marlatt and Donovan, 2007), including AA doctrine. In this regard, recent attention has focused on formally testing how reductions in different aspects of negative affect among AA members, including depression, may account for AA-related benefit (Kelly et al., 2010; Tonigan et al., 2013; Worley et al., 2012). Third, 12-step treatments [e.g. Twelve Step Facilitation (TSF)], which are formal treatments aimed to increase AA participation and facilitate progression through the 12 steps, are the most popular therapeutic approaches in the United States for the treatment of alcoholism, and active facilitation into AA during and after treatment is a primary factor accounting for the effectiveness of such 12-step-oriented treatment approaches (Tonigan, 2005). Given the high rates of clinical depression among treatment-seeking substance abusers, investigators have therefore focused on the influence of AA participation on clinical depression (e.g., Granholm et al., 2011; Bogenschutz et al., 2013). Collectively, then, the identification of linkages between specific AA-related practices and later reductions in clinical depression has considerable value.

In general, AA attendance is associated with reductions in depression (Kelly et al., 2010; Worley et al., 2012). The causal pathway producing these changes is unclear, however. For example, Kelly and colleagues (2010) found that frequency of AA attendance predicted reductions in depression as assessed by the Beck Depression Inventory (BDI; Beck et al., 1961) over a 12-month period in the aftercare and outpatient samples in Project MATCH. However, Kelly et al. found that this association diminished to non-significance when controlling for concurrent alcohol use. They concluded that the apparent association between AA exposure and depression was explained, in large part, by AA’s direct effect on drinking reduction which, in turn, mobilized reductions in depression. Worley et al. (2012) likewise reported that frequency of AA attendance predicted later reductions in depression as measured by the Hamilton Depression Rating Scale (HAM-D, Hamilton, 1960) Unlike Kelly et al., Worley and colleagues reported that AA attendance had a beneficial effect on depression above and beyond that explained by drinking. Specifically, they reported that the association between AA attendance and reductions in depression remained even after the effects of concurrent alcohol and illicit drug use on depression had been statistically controlled. How can these findings be reconciled? Is it not enough to know that, in general, AA referral and attendance are associated with reductions in depression?

Clarifying the nature of the linkage between AA attendance and subsequent reductions in depression is important for the following reasons. If AA attendance is found to decrease depression above and beyond its effect on drinking, then specific AA-prescribed behaviors and beliefs can be identified in community-based AA that mobilize reductions in depression. The identification of these 12-step practices would be highly informative to addiction and mental health treatment providers when treating individuals in early recovery who are suffering from depressed mood. In contrast, if the linkage between AA attendance and changes in depression is fully explained by reductions in drinking, then providers may be more inclined to focus efforts to improve mood by primarily encouraging AA-related practices with documented benefits on alcohol consumption such as AA attendance and acquiring an AA sponsor (Emrick, Tonigan, Montgomery, & Little, 1993; Kelly & Moos, 2003; Tonigan & Rice, 2010).

The purpose of the present study was to rigorously investigate the nature of the linkage between 12-step attendance and reductions in depression. Unique to this study, we recruited only participants early in the AA affiliation process who did not have extensive AA histories to control for past learning effects. Second, it is common for individuals to attend AA and formal treatment concurrently. Investigations of AA efficacy rarely control for the effects of such treatment exposure, however. Planned analyses in this study explicitly controlled for treatment exposure. Third, the sample was followed for a longer period of time (24 months) than previous studies of AA and depression allowing for exploration of effects of long-term AA exposure. With this background, this study had three aims. First, we sought to replicate the finding that depression scores declined over time and that concurrent alcohol use accounted, in part, for this reduction. Our second aim focused on clarifying the lagged association between AA attendance and depression while controlling for concurrent drinking and treatment. Finally, our third aim sought to identify AA-related practices (spirituality, step work) that prospectively predicted changes in depression beyond that accounted for by attending formal treatment, AA, and drinking reductions.

Method

Participants and Procedure

Participants were part of a large sample recruited for the purpose of studying mechanisms of behavior change associated with AA (R01-AA014197). Two hundred fifty three adults with alcohol dependence were recruited from AA groups (n=68), from outpatient substance abuse treatment facilities (n=87), or from community sources including homeless shelters, advertisement in neighborhood newspapers, and flyers (n= 98). To recruit adults early in their exposure to AA, participants were excluded if they had more than 16 weeks of lifetime AA exposure or if they reported having achieved a period of alcohol abstinence of at least 12 months at any time in their life after their alcohol use had become a problem. In addition, participants were required to have attended one or more AA meetings in the prior 3 months. Participants were also required to have consumed alcohol in the prior 90 days, and to meet Diagnostic and Statistical Manual of Mental Disorders (4th ed., DSM-IV; American Psychiatric Association, 2000) criteria for alcohol dependence or abuse (SCID; First, Spitzer, Gibbon, & Williams, 2002). Of the 253 participants, about 66% of the sample was male (n = 168) and the mean age of participants was 38.62 years (SD =9.78). A majority of the sample had a high school degree (20.7% no degree, 14.7% GED, 38.6% high school diploma), was unemployed (64.5%) and were single or divorced (42.0%, 32.8%). 17.2% of the participants reported being homeless. Of the sampled participants, 40.3% were Hispanic, 34.5% were non-Hispanic White, 17.1% were Native American, and the remaining participants were of African, Asian, or unspecified ancestry. All procedures were approved by the institutional review board at the University of New Mexico (UNM Protocol No. 24028).

All participants completed semi-structured interviews and questionnaires at baseline, 3, 6, 9, 12, 18 and 24 months, most often in person, but occasionally by telephone if needed. If participants missed an interview, but were then interviewed later, data from the missed time point were reconstructed. In brief, depending on the time point, 75–85% of the interviews were contemporaneous (performed at the appropriate time point) and 9.9–19% were reconstructed (for further details, see Jenkins & Tonigan, 2011). Both contemporaneous and reconstructed data were used for this study, which was determined to be valid for this dataset in previous work (Jenkins & Tonigan, 2011). More than 90% of the original sample provided follow-up data at 18 months and more than 85% of the original sample at 24 months (see Table 1 for details).

Table 1.

AA Attendance, Spirituality, Drinking Behaviors, and Depression Baseline Through 24 Months

	Months from Baseline
Measure	0	3	6	9	12	18	24
Meeting Attendance*, M (SD)	0.17 (0.18) n=253	0.29 (0.32) n=239	0.19 (0.25) n=239	0.15 (0.23) n=237	0.15 (0.24) n=233	0.14 (0.23) n=229	0.12 (0.21) n=217
% Attending a Meeting During Interview Period	100	79.9	69.5	62.4	59.2	58.1	61.8
Treatment*, M (SD)	0.08 (0.15) n=252	0.12 (0.22) n=238	0.10 (0.25) n=239	.06 (.11) n=237	0.05 (0.14) n=233	0.06 (0.16) n=229	0.06 (0.21) n=217
% Attending a Treatment Session During Interview Period	57.1	61.9	52.3	46.5	38.2	45.4	41.9
Percent Days Abstinent, M (SD)	53.80 (30.42) n=253	77.19 (33.55) n=239	79.54 (31.11) n=239	77.85 (33.18) n=237	78.20 (32.71) n=233	79.41 (32.64) n=229	79.43 (31.55) n=217
Drinks Per Drinking Day, M (SD)	17.74 (12.67) n=253	8.26 (10.65) n=239	7.21 (8.67) n=239	6.77 (8.46) n=237	6.80 (8.71) n=233	6.81 (8.71) n=229	6.58 (8.96) n=217
RBB*, M (SD)	19.15 (8.87) n=253	20.69 (8.81) n=219	19.90 (9.62) n=215	21.29 (9.72) n=202	21.17 (9.70) n=213	21.05 (9.98) n=215	21.66 (10.06) n=216
Surrender Steps*, M (SD)	1.53 (1.30) n=246	0.99 (1.12) n=220	0.66 (0.98) n=215	0.59 (0.95) n=202	0.58 (0.93) n=213	0.50 (0.89) n=207	0.54 (0.96) n=215
Action Steps*, M (SD)	0.93 (1.73) n=246	0.38 (0.97) n=220	0.46 (1.09) n=215	0.53 (1.17) n=202	0.45 (1.05) n=213	0.42 (0.95) n=207	0.56 (1.30) n=215
Maintenance Steps*, M (SD)	0.30 (0.79) n=246	0.09 (0.43) n=220	0.10 (0.51) n=215	0.18 (0.64) n=202	0.13 (0.52) n=213	0.11 (0.45) n=207	0.14 (0.56) n=215
Total BDI* Score, M (SD)	19.79 (11.35) n=227	16.15 (10.97) n=205	14.99 (11.60) n=215	miss	13.30 (10.58) n=213	13.15 (11.53) n=215	12.58 (11.82) n=215

Note. Statistics were calculated using all available data from each interview.

^*Meeting Attendance = proportion of days of 12-step attendance for each participant (# days meeting attended/# days in assessment period, means included zero). Treatment = proportion of days of self reported days of formal therapy (residential and/or outpatient for emotional, drug or alcohol problems) for each participant (# days in which treatment attended/# days in assessment period, means included zero). RBB = Religious Background and Behavior questionnaire. Surrender Steps: Range 0–3 Action Steps: Range 0–6, Maintenance Steps: Range 0–3. BDI = Beck Depression Inventory.

Measures

Although a broad array of data was collected in this study, we will describe only those measures which are relevant to the aims of the present study.

Alcohol use

Retrospective reports of alcohol use were collected using the Form 90 (Miller, 1996), a calendar-based structured interview. Two alcohol use measures from the Form 90 were computed. Proportion of days abstinent from alcohol (PDA) was defined as the number of alcohol-abstinent days in an assessment period divided by the total number of days in the period. Drinks per drinking day (DPDD) was defined as number of drinks consumed per drinking day divided by the number of drinking (i.e., non-abstinent) days in the assessment period.

Help seeking

A single item from the Form 90 interview was used to document the frequency of 12-step meeting attendance during an interview period. Because the exact number of days in an interview period varied slightly between participants, we calculated the proportion days of 12-step attendance for each participant (i.e., days 12-step meeting attendance divided by the number of days in an interview period). Concurrent treatment attendance (Treatment) was also obtained from the Form 90, and included self reported days of formal therapy (residential and/or outpatient for emotional, drug or alcohol problems). Using the method just described for the 12-step meeting attendance, we computed the proportion days of formal treatment for each person at each interview.

Spirituality and religiousness

Spirituality and religiousness was assessed with the Religious Background and Behavior questionnaire (RBB; Connors et al., 1996). This is a 13-item multidimensional scale that measures self-reported degree of belief in God, on a single question (5-point Likert scale), as well as lifetime, and past year, engagement in religious practices. We focused on past year engagement in religious practices (excluding the lifetime practices) which, in addition to the belief in God question mentioned above, asked 6 additional questions on an 8-point Likert scale regarding how frequently over the last 90 days participants 1. Thought about God, 2. Prayed, 3. Meditated, 4. Attended worship services, 5. Read or studied scriptures/holy writings, 6. Had direct experiences of God.

12-step Step work

The Alcoholics Anonymous Involvement questionnaire (Tonigan, Connors, & Miller, 1996) was used to measure step work which was grouped as Surrender Steps, Action Steps and Maintenance Steps. Specifically, participants were asked if, during the assessment period, they had completed any of the 12 steps. For each ‘yes’ answer, they were given a score of 1. The value for the Surrender Steps variable ranged from 0–3, and reflected the number Steps 1, 2, or 3 the participant reported they had completed during the assessment period. The value for the Action Steps variable ranged from 0–6 and reflected the number of Steps 4, 5, 6, 7, 8 or 9 the participant reported they had completed during the assessment period. The value for the Maintenance Steps variable ranged from 0–3 and reflected the number of Steps 10, 11, and 12 that the participant reported they had completed during the assessment period.

Depression

The Beck Depression Inventory II (BDI; Beck, 1996) was used to measure severity of depression. This is a widely used questionnaire used in clinical and research settings. This questionnaire is composed of 21 questions scored on a scale of 0 to 3 asking about depression symptoms over the past 2 weeks such as hopelessness, irritability, guilt, feelings of being punished, fatigue, weight loss, and lack of interest in sex. Total scores of 0–13 indicated minimal depression, 14–19 mild depression, 20–28 moderate depression and 29–63 severe depression.

Statistical Analyses

For Aim 1, 2 and 3 we used hierarchical linear modeling (HLM; Raudenbush & Bryk, 2001) and, in particular, we used a lagged growth model. Analyses were performed with HLM 6 (Raudenbush, Bryk, & Congdon, 2004), and used restricted maximum likelihood estimation. Notably, the dataset was missing BDI scores at 9 months, although we did have data for the other measures at this time-point. In all cases time was coded as 0, 3, 6, 12, 18, and 24 months following baseline and entered as a fixed factor.

Aim 1: To determine whether depression decreases over 2 years of AA exposure, and whether or not alcohol use accounts, in part, for these decreases

For this aim, BDI was assigned as the outcome variable, and concurrent PDA, DPDD (fixed) were entered as level 1 predictors.

Aim 2: To clarify the prognostic value of AA attendance on depression while controlling for concurrent drinking and treatment followed by formal tests of mediation

For this aim, AA attendance (lagged) and time-varying covariates DPDD, PDA and Treatment (same time-points as BDI) were entered as level 1, fixed independent variables, and BDI was assigned as the outcome variable (Figure 1). Because the dataset was missing BDI scores at 9 months, we used the following time-lagged ordering of the dataset for the lagged pairs: baseline AA attendance was used to predict BDI scores at 3 months, AA attendance at 3 months to predict BDI scores at 6 months, AA attendance at 9 months to predict BDI scores at 12 months, AA attendance at 12 months to predict BDI scores at 18 months, and AA attendance at 18 months to predict BDI scores at 24 months. To examine whether the effect of AA attendance on depression could be partially or fully accounted for by alcohol use (DPDD or PDA) we conducted several mediation tests in a smaller sample of participants (those without missing data; n=208) using the bias-corrected bootstrap (Efron & Tibshirani, 1993) as implemented in the PROCESS macro for SPSS (Hayes, 2013). The bias-corrected bootstrap does not rely on the tenuous assumption that the indirect effect is normally distributed and has been shown to be a powerful test of mediation (Fritz & MacKinnon, 2007). In all models, we bootstrapped 50,000 samples and significant effects were determined by 95% bias-corrected confidence intervals that do not contain zero. To maximize statistical power, we conducted two single-mediation models in which we examined the total, direct, and indirect effects of AA attendance (3 months) on BDI scores (24 months) via PDA and DPDD (12 months) while controlling for covariates (baseline DPDD, PDA, and treatment attendance).

Figure 1.

Pictoral representation of statistical model used for Aim 2. All variables were fixed. (DPDD = drinks per drinking day. PDA = percent days abstinent. Tx = treatment. AA = 12 step attendance. BDI = Beck Depression Inventory.)

Aim 3: To identify the prognostic value of AA-related practices (spirituality, step work) on depression beyond that accounted for by attending formal treatment, AA, and drinking reductions followed by formal tests of mediation

For Aim 3 we also used a lagged growth model, which was constructed in a similar fashion as Aim 2. As in Aim 2, BDI was entered as the dependent variable, and AA attendance (lagged) and time-varying covariates DPDD, PDA and Treatment (same time-points as BDI) were entered as independent variables. However, in this model, we added in four additional fixed level 1 predictors, in a time-lagged fashion. These predictors were RBB, Surrender Steps, Action Steps and Maintenance Steps (Figure 2). To examine whether the effect of AA attendance on depression could be partially or fully accounted for by AA-related practices (RBB, Step Work) we conducted several more mediation tests on a smaller sample (those without missing data; n=197), as was done in Aim 2. We conducted four single-mediation models in which we examined the total, direct, and indirect effects of AA attendance (3 months) on BDI scores (24 months) via one of four mediators: RBB, Surrender Steps, Action Steps, and Maintenance Steps (12 months), while controlling for covariates (baseline DPDD, PDA, and treatment attendance).

Figure 2.

Pictoral representation of statistical model used for Aim 3. All variables were fixed. (DPDD = drinks per drinking day. PDA = percent days abstinent. Tx = treatment. AA = 12 step attendance. RBB = Religious Background and Behavior questionnaire. St-Surr = surrender steps. St-Act = action steps. St-Main= maintenance steps. BDI = Beck Depression Inventory.)

Results

Table 1 provides descriptive information from baseline through the 24 month follow-up for relevant measures such as DPDD, PDA, AA attendance, RBB scores, step work, and BDI.

Aim 1:To determine whether depression decreases over 2 years of AA exposure, and whether or not alcohol use accounts, in part, for these decreases

Using HLM 6 (Raudenbush, Bryk, & Congdon, 2004), we conducted a random intercept model with BDI scores as the outcome variable and found that there was sufficient variability in BDI scores to partition within-subject and between-subject effects (ICC = 0.519). Next, we added time as a fixed factor (centered at baseline and coded as 0, 3, 6, 12, 18, and 24 months following baseline) and found that there was a significant linear decrease in BDI scores over time (t = −9.16, b =−0.26, p < 0.001; g = 0.57). Third, we added time as a random factor and found there was significant heterogeneity in the rate of change over time, χ²(240) = 312.09, p < 0.001. Finally, we added DPDD and PDA as level 1 predictors (fixed) and found that increased abstinence (PDA) was associated with lower depression (t = −6.83, b = −7.08, p < .001; g = 0.43) whereas increased drinking intensity (DPDD) was associated with higher depression (t = 2.78, b = 0.08, p = 0.006; g = 0.17) (Table 2). Collectively, these three predictors accounted for 11% of the variance in changes in BDI scores over time and in reference to the unconditional model, the full model provided a significantly better fit to the observed changes in BDI scores, χ²(2) = 200.18, p < 0.001.

Table 2.

Aim 1 – Concurrent Drinking Predicting Depression

	t	Slope (b)	p
Time	−9.16	−0.26	<0.001
DPDD	2.78	0.08	0.006
PDA	−6.83	−7.08	<0.001

DPDD = drinks per drinking day. PDA = percent days abstinent. AA = 12 step attendance. RBB = Religious Background and Behavior questionnaire. St-Surrender = surrender steps. St-Action = action steps. St-Mainenance = maintenance steps.

Figure 3 displays mean BDI scores at each time point for four groups defined by an individuals’ baseline BDI score (N = 227). The four BDI groups were defined as follows: minimum (0–13), mild (14–19), moderate (20–28), and severe (29–63). All available data were used for post-baseline mean time-points. Notably, baseline PDA was significantly (p < 0.01) different among these four groups such that highest BDI severity (severe) had the lowest PDA (0.43) and the lowest BDI group (minimum) had the highest PDA (0.62). Descriptively, decreases over time in mean BDI scores were largest for those individuals with the most severe depression, and at earlier time-points. However, it declined in all four subgroups, and continued to decline even at later time-points.

Figure 3.

Depicts the change in depression (Beck Depression Inventory score) over time for individuals with minimum (0–13), mild (14–19), moderate (20–28), and severe (29–63) levels of depression at baseline assessment.

Aim 2: To clarify the prognostic value of AA attendance on depression while controlling for concurrent drinking and treatment followed by formal tests of mediation

Controlling for concurrent treatment attendance and drinking (DPDD and PDA), AA attendance significantly predicted later decreases in BDI scores (t = −2.80, b = −3.40, p = 0.006; g = 0.18). Although concurrent treatment was not significantly associated with BDI scores (t = 0.04, b = 0.04, p < 0.97; g = < 0.01), concurrent DPDD (t = 3.15, b = 0.11, p = 0.002; g = 0.20) and PDA (t = −4.84, b = −5.76, p < 0.001; g = 0.38) both predicted BDI scores such that more drinking was associated with higher BDI scores. Time continued to be a significant predictor (t = −5.73, b = −0.18, p < 0.001; g = 0.17) (Table 3). In a post hoc analysis, we entered AA as a random factor and found that the individual estimates of the lagged association between AA attendance and BDI scores was homogeneous (χ² (229)= 263.35, p=0.06). Compared to the baseline random intercept model, the full model tested in Aim 2 accounted for 15% of the variance in BDI scores and provided a significantly better fit, χ²(5) = 127.87, p < 0.001.

Table 3.

Aim 2 – Lagged AA Attendance Predicting Depression, Controlling for Concurrent Drinking and Treatment

	t	Slope (b)	p
Time	−5.37	−0.18	<0.001
DPDD	3.15	0.11	0.002
PDA	−4.84	−6.12	<0.001
Treatment	0.04	0.04	0.97
AA	−2.80	−3.40	0.006

DPDD = drinks per drinking day. PDA = percent days abstinent. AA = 12 step attendance. RBB = Religious Background and Behavior questionnaire. St-Surrender = surrender steps. St-Action = action steps. St-Mainenance = maintenance steps.

Post hoc analyses were done to determine if severity of baseline BDI scores moderated the lagged association between AA attendance and BDI scores during follow-up. As was true for Figure 1 (Aim 1), only 227 participants were included in this analysis, as only this number of participants had provided complete baseline BDI data. For this analysis, BDI was divided into two categories: minimum and mild were called low BDI and assigned a zero, moderate and severe were called high BDI and assigned a 1, and BDI category was entered as a level 2 predictor. The cross level interaction was not significant, suggesting that declining rates of BDI scores associated with AA attendance in this sample did not significantly differ between participants initially reporting higher or lower BDI scores at baseline. Specifically, the association between AA attendance and later BDI scores for the low BDI group was b = −4.34 and was b = −1.06, for the high BDI group.

In both models for the mediation analyses, we found significant total (b = −7.55) and direct effects (−6.90 < bs < −6.43) of AA attendance on BDI scores. PDA but not DPDD had a significant indirect effect, indicating that AA attendance was associated with increased abstinence, which was in turn related to decreased depression (b = −1.13, 95% CI: −2.74, −.06).

Aim 3: To identify the prognostic value of AA-related practices (spirituality, step work) on depression beyond that accounted for by attending formal treatment, AA, and drinking reductions followed by formal tests of mediation

When RBB and the step variables (Surrender Steps, Action Steps, and Maintenance Steps) were added into the model, the effect of AA attendance was not significant (t = −1.94, b = −2.25, p = 0.052; g = 0.07). Also, as previous, concurrent Treatment was not a significant predictor of BDI, whereas concurrent drinking as measured by DPDD and PDA and Time continued to be significant predictors. Moreover, RBB (t = −2.27, b = − 0.10, p = 0.02; g = .008) and Action Steps (t = 2.13, b = 0.53, p =0.03; g = 0.07) were significant predictors of later BDI scores such that greater spirituality predicted lower BDI scores, and greater number of Action Steps taken predicted higher BDI scores. However, Surrender Steps and Maintenance Steps were not significant predictors of later BDI scores (Table 4). Compared to the baseline random intercept model, the full model tested in Aim 3 provided a significantly better fit, χ²(5) = 242.00, p < 0.001.

Table 4.

Aim 3 – Lagged AA Attendance, Step Work & Spirituality Predicting Depression, Controlling for Concurrent Drinking and Treatment

	t	Slope (b)	p
Time	−4.88	−0.18	<0.001
DPDD	3.35	0.13	0.001
PDA	−5.11	−6.10	<0.001
Treatment	−0.30	−0.31	0.77
AA	−1.94	−2.25	0.05
RBB	−2.27	−0.10	0.02
St-Surrender	0.23	0.06	0.82
St-Action	2.13	0.53	0.03
St-Maintenance	−1.24	−0.65	0.22

DPDD = drinks per drinking day. PDA = percent days abstinent. AA = 12 step attendance. RBB = Religious Background and Behavior questionnaire. St-Surrender = surrender steps. St-Action = action steps. St-Mainenance = maintenance steps.

We then used partial correlations to explore the relationship between RBB and BDI (time-lagged), controlling for concurrent PDA, DPDD, Treatment, and time-lagged AA attendance. The strength of the association between earlier RBB and later BDI scores were as follows (zero-order Pearson correlation coefficients): 0–3 month=0.05, 3–6 month=−0.03, 9–12 month=−0.06, 12–18 month=−0.12, 18–24 month=−0.11; ps>0.05), indicating that the strength of the relationship between RBB and later BDI scores could be greater at later time points.

In all models for the mediation analyses (RBB, Surrender Steps, Action Steps, Maintenance Steps), we found significant total (b = −8.55) and direct effects (−8.59 < bs < −7.78) of AA attendance on BDI scores. None of the variables had a significant indirect effect.

Discussion

This study investigated the relationship between AA attendance and depression in individuals new to AA. Findings indicated that depression decreased over time during early AA exposure and that reductions in alcohol consumption only partially explained the observed changes in depression. Study findings also provide considerable currency to the idea that AA attendance and AA-related practices predict reductions in depressive symptoms beyond that explained by increased abstinence. Consistent with Kelly et al. (2010), our findings support the existence of a prospective association between AA attendance and later depression via its association with alcohol consumption. Specifically, mediation analyses for PDA but not DPDD supported an indirect effect of AA on depression via drinking. However, unlike Kelly et al. (2010), we also found evidence that AA attendance has beneficial effects on later depression above and beyond its effect on drinking (i.e., a significant lagged effect of AA attendance on depression, even when controlling for concurrent drinking), which was consistent with more recent work (Worley et al., 2012). One important difference between our study and the Kelly et al. (2010) study is that our participants were new to AA, whereas those in Kelly et al. (2010) were not all new members. Early AA affiliates may have a greater magnitude of reduction in depression symptoms early on due to the effect of novelty, in a similar way that placebo pills can induce improvement in depression (Buchanan & Bardy 2010; Walsh et al., 2002). Also notable in our study, the beneficial lagged association between AA attendance and depression (after controlling for drinking) was observed regardless of the severity of their baseline depressive symptoms. It appears, then, that there is a linkage between AA exposure and psychological improvement in early AA affiliates above and beyond its effect on drinking, and that this association may apply to clinically depressed adults as well as to adults with sub-threshold levels of depressive symptoms.

The salutary effect of AA attendance on depressive symptoms is important because comorbid major depression is associated with a variety of other problems among individuals with substance use disorders including decreased rates of remission and increased rates of relapse (Hasin et al., 2002) as well as elevated rates of suicide attempts (Bolton et al., 2010). Interestingly, our findings suggest that the decreases in depression associated with AA participation is initially rapid but that it also continues, on average, throughout 24 months of AA attendance. Continuing to attend AA and to engage in AA-related practices related to spiritual growth in particular may provide psychological benefits above and beyond beneficial effects on alcohol consumption months to years after initial exposure.

This study provided two analytical perspectives to judge changes in depressive symptoms among adults with AUD trying to reduce their drinking. Ignoring AA attendance patterns, Figure 1 indicated that reductions in symptom counts were largest and most rapid for those participants who, at baseline, were in the most severe range on the BDI. Inspection of baseline alcohol use for this group suggests that these reductions may have been associated with higher rates, on average, of alcohol use. In contrast, findings indicated that the benefits of AA meeting attendance on reducing later depressive symptom counts was not moderated by baseline depression severity. Here, the lagged association between AA attendance and depressive symptom counts was aggregated or summarized across all time points in investigating the moderating effects of baseline depression, and alcohol use was taken into account. Although the former perspective might imply that AA has greater effect on depressive symptoms in those with higher levels of depression, the latter perspective provides convincing evidence that above and beyond effects on drinking, AA has psychological benefits even in those with minimal depression.

Several prospective longitudinal studies have now reported that gains in spiritual practices accounted, in part, for increased abstinence among AA members (Connors et al., 2001; Kaskutas et al., 2003; Kelly, Stout, Magill, Tonigan, & Pagano, 2011; Krentzman, Cranford, & Robinson, 2013; Robinson, Krentzman, Webb, & Brower, 2011; Tonigan, Rynes, & McCrady, 2013; Zemore, 2007). Using formal mediational tests, some of these studies have demonstrated first that frequency of AA meeting attendance predicted increased spiritual practices, and that such gains predicted later decreases in alcohol consumption (usually abstinence) (Kelly et al., 2012; Krentzman et al., 2013; Tonigan et al., 2013; Zemore, 2007). Our findings from the mediation analysis and lagged analysis suggest that spiritual gains may indeed predict later reductions in depressive symptom counts among AA members, even though engagement in spiritual practices are not described by AA attendance frequency. Engagement in spiritual practices may improve well-being by increasing self-esteem, improving stress-coping, providing meaning, and decreasing social isolation (Beshlideh, 2009; Koenig, 2009). Although AA attendance has been found to predict both significant increases in spirituality and decreases in depression (Kelly et al., 2012), our mediation analysis did not demonstrate an indirect effect of AA attendance on depression; however we necessarily did the mediation analysis in a smaller sample of participants (only those with complete data), which may have obscured true findings. Our mediation analysis was also unorthodox in that the time points between variables were unusually far apart (3 months, 12 months, and 24 months), making it more difficult to achieve significance. Moreover, that AA attendance was no longer significant in Aim 3 after adding spirituality (and step work) variables into the model implies that these variables may be accounting for a large portion of the variance for the association between earlier AA attendance and later depression. To our knowledge, this finding offers the first demonstration that this evidence-based mechanism of change (spirituality) in community-based AA may have multiple beneficial actions.

The finding that extent of self-reported completion of steps 4–9 (Action Steps) was positively associated with later depressive symptom counts was perplexing. In particular, it is well-documented that having an AA sponsor is predictive of alcohol abstinence (Bond et al., 2003; Witbrodt & Kaskutas, 2005), especially during early AA affiliation (Tonigan & Rice, 2010). A primary role of an AA sponsor is to guide a new member through the 12-steps, and evidence suggests that AA members with sponsors are significantly more likely to endorse the practices and precepts in steps 4–9 relative to AA members without a sponsor (Greenfield & Tonigan, 2013). We offer some possibilities for our unexpected finding. First, self-reported completion of an AA step is laden with measurement challenges and our measures of step work may have lacked sufficient reliability. To report the completion of a first step requires a subjective evaluation that is made all the more difficult because each of the 12-steps contain multiple prescriptions. These measurement challenges are further exacerbated by the stated conventional wisdom voiced in 12-step meetings that step work is a life-long process and that steps are, in reality, never completed. Recent work suggests that the procedure of directly asking respondents which steps they have completed (as done in this study) can lead to unacceptably high rates of false negatives relative to asking AA members the extent that they endorse specific prescriptions in each of the 12-steps (Greenfield & Tonigan, 2013). Ultimately, then, our finding may simply be a measurement artifact. A rival explanation can also be offered. Specifically, the focus of the action steps is on compiling a moral inventory, the sharing of this inventory, and the making of amends for past transgressions identified in an inventory. While these actions may enhance the reduction of alcohol use they may also evoke feelings of guilt and remorse which become manifest in higher depressive symptom counts. Also notable was that our mediation analyses (albeit in a smaller number of participants and over a longer spread of time) showed a large negative association between action step work and depression.

That step work in general was not predictive of lower depressive symptoms is also notable, and adds to the literature exploring the importance of step work (relative to other AA associated-behaviors such as getting a sponsor, meeting attendance, socialization spirituality etc.) in the beneficial effect of AA on recovery. Whether or not step work is a key mechanism by which AA improves drinking outcomes is unclear, and results are mixed (Cloud, Ziegler, & Blondell, 2004; Zemore, Subbaraman, & Tonigan, 2013). Moreover, although associations between overall step work and psychosocial benefits (interpersonal insecurity, social potency) have been observed (Suire and Bothwell, 2006), to our knowledge there has not been a study investigating the lagged relationship between step work and mood. Our results do not lend support to hypotheses that step work is a key ‘magic ingredient’ for psychological well-being during recovery. Clearly, our findings in this area raise more questions than point to answers and we encourage future research on this topic.

Our findings have important implications for clinicians involved in treating individuals with AUD. In addition to AA attendance, practices to enhance spirituality could also be encouraged, with such recommendations having potential to benefit both alcohol use and mood. Our results also support that both AA attendance and practices to enhance spirituality should be encouraged as integral parts of achieving and maintaining alcohol abstinence among individuals with AUD, and may continue to provide lasting benefits as far out as 2 years from the first AA meeting. Unfortunately, besides encouraging AA attendance, it is not yet clear what the best specific recommendations and interventions should be to help individuals enhance spiritual growth. Although religion-based psychological interventions may work at least as well as (Hook, Worthington Jr., Davis, Jennings II, Gartner, & Hood, 2010) or better than standard therapies for depression (Koenig, 2009), findings from work integrating therapeutic strategies to increase spirituality into standard SUD treatment have been mixed, with a 7-week ‘Knowing Your Higher Power’ behavioral intervention showing promise in 12 step recovery based treatment (Brown et al., 2007), but no improvement in drinking measures and increases in anxiety and depression being observed when treatment as usual was augmented with a spiritual intervention by a professional spiritual director (Miller, Forcehimes, O’Leary, & LaNoue, 2008). However, in this same study, increases in anxiety and deprssion were not observed when the spiritual intervention was delivered by a treatment provider (Miller et al., 2008). These findings, coupled with evidence that in other marginalized populations (e.g. pregnant teenagers) religiousness is associated with greater depression scores (Koenig, 2009), indicate that stigma may account for some of the negative effects of religiousness and spirituality on mood in some populations. Whoever is delivering the spiritual intervention should ideally be accustomed to treating the population in question, and be able to do so in a non-judgmental manner.

In spite of the significant strengths of this study it is important to note several limitations. Individuals with extensive AA attendance histories were excluded in an effort to minimize the potential confound of past AA learning effects on the relationships of interest. This exclusion may limit the generalizability of our findings given that adults with AUD frequently cycle into, through, and out of AA. Similarly our sample was 40% Hispanic, and, although this is representative of the New Mexico population, our findings may not be generalizable to other regions with different ethnic distributions. We were unable to conduct non-response analyses as we only had data from individuals consented who followed through. Follow up rates were quite good (> 90% at 18 months, and >85% at 24 months), so attrition analyses were not conducted. Second, the Form 90 interview used to derive the frequency of AA attendance variable used in this study asked about frequency of 12-step meeting attendance. We strongly suspect that AA was the predominant type of 12-step meeting attended by our participants because all study participants were recruited based upon their attendance at AA meetings in the 90-day period before study enrollment. Nevertheless, migration across 12-step sister programs is relatively common and it likely that some non-AA 12-step attendance was incorrectly classified as AA attendance. Third, the tool used in this study to measure religious and spiritual beliefs and practices is widely used in 12-step research (Connors et al., 1996) although it is not specifically tailored to measure the construct of spirituality as it is practiced in community-based AA. In this light, our study may have provided a conservative “signal” in studying the relationship between spiritual growth in the context of AA and changes in AA member depression. Fourth, self-selection biases such as reporting biases could certainly be a factor driving our associations (eg. those that have a higher need to please the researcher, may be more likely to report AA attendance and spiritual gains as well as lower levels of depression). Unfortunately, in this dataset, we did not have a social desirability measure (or measures for other variables which could be biasing results such as coping, hope, physical functioning, therapeutic alliance etc.) to control for such effects. However, this issue of self-selection bias in 12 step research has fueled a handful of studies, and only one (Fortney 1998) out of four (Humphreys, Phibbs & Moos, 1996; Magura, McKean, Kosten & Tonigan 2012; Yu & Kaskutas, 2009) demonstrated that there was evidence of self-selection bias influencing the associations between AA attendance and drinking outcomes. Therefore, we feel reassured that this was unlikely to be driving our results. Finally, we did not ask participants if they were taking antidepressants or undergoing depression-specific psychotherapy, so we were unable to account for the possibility that these factors may have been driving the effects we observed in this work.

In conclusion, our findings support that, although reductions in alcohol consumption partially explains decreases in depression over time in individuals attending AA, other factors such as AA attendance and spirituality are also contributing to decreases in depression above and beyond their effects on drinking. These findings have important implications for treatment providers. In efforts to lessen depression in individuals with co-occurring major depressive disorder (MDD), or to increase overall well-being in those without it, providers should consider recommending AA and/or other behaviors that enhance spiritual growth.