Beyond Attendance: Social Anxiety and Psychiatric Concerns Predict Mutual-Help Group Involvement in a U.S. National Longitudinal Survey

Abstract

Objective:

Involvement in mutual-help groups (MHGs) is associated with positive alcohol recovery outcomes. Attendance is a first step to involvement, but barriers to attendance remain poorly understood, especially among second-wave (non-12-step) MHGs. This study aimed to describe the most common barriers to MHG attendance, describe variation in barriers across diverse MHG types, and identify attendance barrier domains associated with MHG involvement over a 12-month period.

Method:

Data were from the Peer Alternatives (PAL) Study 2021 Cohort, a longitudinal, online survey of second-wave and 12-step MHG participants (n=531) with follow-ups at 6 and 12 months. Surveys measured MHG attendance barriers (11-items, 3 domains) and MHG involvement (5-items). We employed adjusted Generalized Estimating Equations to examine lagged associations between barrier subscale scores and MHG involvement at 6 and 12 months.

Results:

Social anxiety and psychiatric concerns were overall more highly endorsed than low motivation/perceived need and dislike of meeting attendees and content in the total sample and for each MHG except LifeRing. The most highly endorsed individual barrier to attendance was “I don’t like crowds or large groups.” Higher social anxiety and psychiatric concerns domain scores predicted lower MHG involvement at 6- and 12-month follow-ups in adjusted models (β=−0.09 (−0.18, −0.01), p<0.01), and this was the only barrier domain associated with MHG involvement.

Conclusions:

Social anxiety and psychiatric concerns are salient barriers to attending a variety of MHG groups and to MHG involvement. Efforts to enhance MHG social experiences and the availability of groups for people with psychiatric concerns could improve MHG involvement.

INTRODUCTION

Involvement in mutual-help groups (MHGs) represents the depth of MHG engagement and can include activities such as attending meetings regularly, leading them, volunteering at meetings, or taking on service roles. Attendance, measured as the number of meetings attended over a period of time, is often a first step toward involvement, but may not be sufficient to achieve MHG recovery benefits (Kelly et al., 2020; Timko et al., 2024). Involvement, however, is associated with a myriad of positive outcomes. For example, involvement in well-known 12-step groups, such as Alcoholics Anonymous (AA), is associated with improved abstinence rates (Timko, 2008; Tonigan et al., 1996) and reduced alcohol use (Blonigen et al., 2013; Humphreys et al., 2014) and Alcohol Use Disorder (AUD) symptoms (Timko et al., 2000) improved psychological and psychiatric functioning (Kelly, 2022), and lower healthcare costs (Kelly et al., 2020; Mundt et al., 2012). While outcomes of involvement in second-wave MHGs, meaning 12-step alternatives such as Women for Sobriety (WFS), SMART Recovery (SMART), and LifeRing Secular Recovery (LifeRing), are less well studied than for 12-step groups, a growing literature suggests comparably positive outcomes. Cross-sectional and longitudinal studies have found that greater involvement in second-wave MHGs is associated with greater length of sobriety (Atkins & Hawdon, 2007), and comparable AUD recovery outcomes to 12-step groups (Zemore et al., 2018). Notably, attendance in second-wave MHGs may be growing (Bergman et al., 2024), while in recent decades 12-step group attendance has remained stable and high relative to other forms of help-seeking (Zemore et al., 2023). This demonstrates the relevance of both 12-step and second-wave MHGs to the recovery support landscape.

Despite the effectiveness of 12-step MHGs and emerging evidence supporting second-wave MHGs’ effectiveness, only a minority of people with AUDs ever initiate MHG attendance (Timko et al., 2006). A national survey showed that only 4.5% of individuals with past-year AUD and 15.4% with lifetime AUD received support from a 12-step group (Grant et al., 2015). Additionally, discontinuation of MHG attendance is common; for example, 40% of male Veterans Administration patients stopped attending within one year (Kelly & Moos, 2003), and 47% of individuals with AUD who attended AA in their first year post-treatment had discontinued by year eight (Moos & Moos, 2004, 2006). Tsutsumi et al. (Tsutsumi et al., 2020) observed comparable discontinuation rates among second-wave MHGs vs. 12-step groups and that rates of changing from one’s primary group to another MHG group were higher among second-wave MHG than 12-step participants.

Research on barriers to MHG attendance is limited and has focused on 12-step groups. Barriers to 12-step group attendance include discomfort with the spiritual focus (Buxton et al., 1987; Winzelberg & Humphreys, 1999), rigidity regarding total abstinence (Humphreys, 2000), and unsupportive reactions to other psychiatric concerns (Kelly & Yeterian, 2008). However, no known studies have investigated barriers to attending second-wave MHGs. Moreover, no known studies have investigated how barriers to attendance are associated with MHG involvement for any MHG. Identifying which (if any) barriers to attendance are associated with MHG involvement is important because of the benefits of MHG involvement for AUD recovery and the need to identify targets for interventions to increase involvement (Cloud & Kingree, 2008; Kelly & Moos, 2003). Additionally, understanding the differences in attendance barriers across 12-step and second-wave MHGs can inform whether different approaches are necessary to support attendance and involvement in different MHGs.

This study analyzed longitudinal data from a large cohort of people with lifetime AUD already attending 12-step or second-wave MHGs. Aims were to: 1) describe the most common barriers to MHG attendance overall; 2) examine variation in endorsement of attendance barriers across primary group (i.e., the only group or self-reported “primary group” that a participant attended); and 3) identify the barrier domains most predictive of MHG involvement over 12 months. We chose involvement (not attendance) as our main outcome because involvement predicts recovery-related outcomes more strongly than attendance in previous work (Kelly et al., 2013) (Weiss et al., 2005) and in our preliminary analyses as described below. Given the lack of relevant literature and theory, we refrain from providing any specific hypotheses.

METHODS

Sample and procedures

Data were from the Peer Alternatives (PAL) Study 2021 Cohort, a longitudinal, online survey of MHG participants with interviews at baseline, 6 months, and 12 months conducted between 10/20/2021 and 12/5/2022. The study recruited MHG attendees in collaboration with WFS, LifeRing, SMART, IntheRooms (ITR) and Faces and Voices of Recovery (FVR), via email, webpages, social media, flyers, and at in-person and online MHG meetings. Interested parties completed an online screener to determine their eligibility on the PAL Study website. Eligible participants were aged 18+; were U.S. residents; had attended at least one in-person or online meeting of 12-step, WFS, LifeRing, or SMART in the past 30 days; reported lifetime AUD; and had not participated in the PAL Study 2015 Cohort (Zemore et al., 2017). Lifetime AUD was determined using a subset of Composite International Diagnostic Interview (CIDI) items that addressed each of the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) criteria (see Measures) (Zemore et al., 2017).

Data quality review involved two processes to detect fraud and low data quality: 1) an initial review by our data collection partner (ICF Macro, Inc.) evaluating survey meta-data and personal contact information, and 2) an in-depth review by the study team inspecting both meta-data and all survey responses. Best practices informed procedures for ensuring data quality in online surveys (Godinho et al., 2020; Pozzar et al., 2020; Pratt-Chapman et al., 2021; Storozuk et al., 2020). For follow-ups, the cleaned baseline sample received an invitation email including a personalized link to the 6- or 12-month survey. Nonrespondents received up to 2 emails, 2 text messages, a letter with the personalized survey link, and 2 phone calls. Respondents who completed surveys and passed data quality review received Amazon gift codes ($35 at baseline, $40 for each follow-up).

Response rates among those eligible for follow-up after data quality review were 87.8% and 84.6% at 6 months and 12 months, respectively. Participants lost to follow-up at the 6-month survey were more likely to be women (vs. men; OR=1.90, p<0.05), and less likely to report lifetime drug problems (vs. reporting lifetime drug problems; OR=0.55, p<0.05), or SMART (vs. 12-step; OR=0.47, p<0.05) as their primary group at baseline. Similar attrition patterns were observed for the 12-month survey, although LifeRing or SMART attendees were less likely than 12-step to complete the 12-month survey (OR=0.45, p<0.05 and OR=0.43, p<0.05, respectively) (Timko et al., 2024; Zemore, 2024).

The final sample included 531 individuals who had a mean age of 47.8 years, were predominantly female (66.3%), non-Hispanic White (hereafter White) (84.4%), and married/partnered (53.3%). Half (50.3%) reported having a college degree; just under a third reported an income of $40,000 or less (30.4%); and most reported living in an urban or suburban area (81.0%) and being employed (62.2%) (see Table 1).

Table 1. Sample Characteristics by Primary Group at Baseline.

	Total Sample (N=531)	Primary Group
		12-Step (n=105)	WFS (n=131)	LifeRing (n=136)	SMART (n=159)
	%	%	%	%	%
Age
18–29	9.4	17.1	3.8***	11.0**	7.6*
30–39	24.5	35.2	16.8	20.6	27.0
40–49	18.5	20.0	15.3	13.2	24.5
50–59	23.2	15.2	26.7	25.7	23.3
60 or older	24.5	12.4	37.4	29.4	17.6
Gender
Female (vs. male)	66.3	72.4	100.0***	47.0***	50.3***
Race/Ethnicity
Non-Hispanic White	84.4	71.4	87.0**	88.2**	87.4**
Non-Hispanic Black/African American	6.2	16.1	3.8	2.9	4.4
Latinx/Hispanic	5.8	6.7	6.1	5.9	5.0
Other	3.6	5.7	3.1	2.9	3.1
Marital Status
Single	23.0	29.5	9.9***	21.3**	30.8
Separated/widowed/divorced	23.7	29.5	25.2	17.7	23.9
Married/partnered	53.3	41.0	64.9	61.0	45.3
Educational Attainment
Less than college degree	26.7	41.9	16.0***	24.3*	27.7
College degree (AA/BA)	50.3	43.8	47.3	55.2	52.8
Post-graduate training or degree	23.0	14.3	36.6	20.6	19.5
Income
40,000 or less	30.4	43.3	22.3***	23.5**	34.6
40,001 to 80,000	25.5	27.9	23.1	25.7	25.8
80,001 to 120,000	26.1	21.2	30.8	32.4	20.1
More than 120,000	18.0	7.7	23.9	18.4	19.5
Urbanicity
Urban/suburban (vs. rural)	81.0	77.1	73.3	86.8	84.9
Employment Status
Employed full/part-time (vs. unemployed)	62.2	59.1	62.6	60.3	65.4
Lifetime 2+ drug problems
Yes (vs. no)	53.2	61.0	40.8**	47.8*	62.9
Current alcohol recovery goal
Total lifetime abstinence (vs. other)	51.7	50.5	61.1	48.9	47.2
Last alcohol use
1+ years ago (vs. less than 1 year)	40.7	31.4	45.0	37.5	45.9
Mutual-help attendance mode past 30 days
In-person only	12.6	8.6	7.6***	9.6***	22.0***
Online only	53.7	26.7	67.9	72.1	44.0
Both in-person and online	33.7	64.8	24.4	18.4	34.0
	Mean (SE)	Mean (SE)	Mean (SE)	Mean (SE)	Mean (SE)
Lifetime AUD severity	9.70 (0.07)	9.90 (0.15)	9.48 (0.15)	9.63 (0.14)	9.82 (013)
Lifetime psychiatric severity	0.63 (0.01)	0.64 (0.03)	0.64 (0.02)	0.60 (0.03)	0.65 (0.03)
Past 30-day meeting attendance	16.5 (1.0)	26.2 (3.16)	15.1* (1.85)	15.1** (1.67)	12.0*** (1.52)

Note:

^*p-value<0.05

^**p-value<0.01

^***p-value<0.001; ns = not significant; missing values excluded from percentages; p values indicate a significant difference on a given variable for a subgroup compared to 12-step and are presented only given a significant omnibus test with all subgroups.

Measures

MHG attendance variables.

At each survey, participants reported the number of past-30-day in-person and online MHG meetings attended for each MHG type. To code primary group, participants attending multiple MHGs were asked which group they considered their “primary group”; for participants attending only one group, we coded that group as primary. Given eligibility criteria, baseline primary group was limited to MHG target groups (i.e., 12-step, WFS, LifeRing, or SMART). At follow-ups, however, categories included another (non-target) MHG and no MHG. To code past 30-day attendance mode, we created a 3-level categorical variable (in-person, online, or both) collapsing across groups attended.

Barriers to continued mutual-help group attendance.

We used a brief, adapted version of the 24-item REASONS measure (Kelly et al., 2010) to identify barriers to continued MHG attendance at baseline and 6 months. Selected items included the 8 most highly endorsed barriers in prior work on 12-step groups (Kelly et al., 2010); 2 additional items were selected to ensure adequate coverage of critical domains (i.e., 1 item in the motivation/no perceived need domain and 1 item in the psychiatric concerns domain). Minor item modifications included separately inquiring about comfort discussing medication and psychiatric issues, yielding 11 total items. We also changed the item “I don’t like having to speak at meetings” to “I didn’t feel comfortable speaking at meetings” to better reflect contexts of target MHGs. Using a Likert-type scale ranging from 1 (not at all) to 5 (extremely), participants rated the extent to which each of these 11 factors was a barrier to attending their primary group in the prior 30 days. We used Exploratory Factor Analysis (EFA) (see Analyses) to identify subscales and computed average subscale scores (see Table 2 for items).

Table 2. Barrier Scale Item Factor Loadings, Item and Subscale Means, and Subscale Reliability Coefficients at Baseline.

Subscale	Items	Factor 1 Loading	Factor 2 Loading	Item Mean (SD)	Subscale Mean (SD) (Range: 1–5)	Subscale Internal Reliability (α)
Low motivation or perceived need	I wasn’t ready or motivated to change my drinking/drug use.	0.76	−0.14	1.60 (1.02)	1.69 (0.90)	0.76
	I wanted to cut down, but not stop drinking/taking drugs completely.	0.71	−0.08	1.56 (1.03)
	I thought I could change and do it myself.	0.63	0.07	1.90 (1.23)
Dislike of meeting attendees and content	People at meetings were too rigid about things.	0.61	0.09	1.51 (0.94)	1.74 (0.85)	0.79
	I didn’t like hearing the same stories over and over again.	0.58	0.17	1.86 (1.10)
	Meetings were just the same thing over and over again.	0.52	0.18	1.84 (1.01)
Social anxiety and psychiatric concerns	I don’t like crowds or large groups.	−0.18	0.85	2.08 (1.26)	1.80 (0.85)	0.83
	I didn’t feel comfortable speaking at meetings.	0.04	0.67	1.76 (1.06)
	It was hard to connect with people.	0.09	0.61	1.82 (0.99)
	I didn’t feel comfortable talking at meetings about my psychiatric problems.	0.30	0.54	1.75 (1.16)
	I didn’t feel comfortable talking at meetings about my medication issues.	0.23	0.53	1.62 (1.05)

Note: Response option for items were 1. Not at all, 2. A little, 3. Somewhat, 4. Very much, 5. Extremely

Mutual-help group involvement.

All three surveys included 4 yes=1/no=0 items to measure past 30-day MHG involvement. We selected and adapted these four items from the Alcoholics Anonymous Involvement scale (Humphreys et al., 1998) because they were strongly associated with substance use outcomes in previous studies (Zemore et al., 2018; Zemore et al., 2013) and are appropriate for all target MHGs. Items inquired whether participants currently had a regular or “home” group; had at least one close friend or “sponsor” whom they could call on for help when they needed it; convened or led any meetings; and did volunteer work or “service” at a meeting. We created a summary measure of MHG involvement based on 5 items by averaging across the 4 involvement items and an item indicating weekly or more (vs. less than weekly) meeting attendance (6-month alpha=0.70, 12-month alpha=0.72). Participants who did not attend any MHG in the past 30 days at the follow-up surveys received a summary score of zero.

Demographics:

Demographics (see Table 1 for categories), assessed at baseline, included age, gender, race, ethnicity, marital status, educational attainment, income, employment status, and urbanicity. Regarding gender, one person identified as transgender and one as “other;” both were recoded to missing.

Clinical characteristics:

Clinical characteristics assessed at baseline included an adapted version of the alcohol section of the CIDI (World Health Organization, 1993) to assess lifetime alcohol use disorder severity. This version included 18 items addressing the 11 criteria for a DSM-5 AUD diagnosis (American Psychiatric Association, 1993) and an item confirming they experienced these symptom criteria within the same 12-month period. We constructed a variable indicating the number of lifetime AUD criteria. Two yes/no items were used to assess lifetime drug problems: 1. “Were there times in your life when you were often under the influence of drugs in situations where you could get hurt, for example when riding a bicycle, driving, operating a machine, or anything else?,” and 2. “Were there times in your life when you tried to stop or cut down on your drug use and found that you were not able to do so?”. Participants who endorsed both items were coded as having lifetime drug problems (Borges et al., 2015; Saha et al., 2012). We operationalized lifetime psychiatric severity as the average value of 4 dichotomous (yes vs. no) items assessing if the respondent ever 1) received counseling, psychotherapy, psychiatric visits for a mental health problem, was ever 2) prescribed medication for a mental health problem, 3) hospitalized for a mental health problem, and 4) diagnosed with a mental health disorder. Values ranged from 0.25 to 1. We assessed current alcohol recovery goal using a single item (Hall et al., 1991) that asked respondents to select one of five recovery goals that was most true for them, ranging from total lifetime abstinence to controlled alcohol use. We dichotomized responses into total lifetime abstinence (“I want to quit using alcohol once and for all, to be totally abstinent, and never use alcohol ever again for the rest of my life”) vs. other. We assessed recency of last alcohol use by asking respondents, “When was the last time you had any alcohol?”, with response options ranging from within the past 30 days to more than 10 years ago. We recoded responses into 1+ years ago vs. less than 1 year ago.

Analyses

Preliminary analyses included descriptive statistics on sociodemographic, clinical, and MHG characteristics overall and by primary group. They also included an examination of the adapted, 11-item REASONS scale’s factor structure using EFA with iterated principal factor extraction and an oblique (promax) rotation (Costello & Osborne, 2005) and Cronbach’s alpha calculation to develop barrier subscales. Additionally, to empirically support the selection of MHG involvement vs. attendance as the outcome, we examined bivariate associations between MHG involvement and attendance at each wave using Pearson’s correlation coefficients, and conducted unadjusted and adjusted lagged Generalized Estimating Equations (Liang & Liu, 1991) to compare the predictive strength of MHG attendance vs. involvement for past 30-day alcohol outcomes.

To describe the most common barriers to MHG attendance (Aim 1) and examine variation in barriers across primary group (Aim 2), we calculated barrier item and subscale descriptives (means, standard deviations) overall and by primary group, and subscale α’s. To test for differences in barrier subscale means across primary group, we used one-way analyses of variance with Bonferroni multiple-comparison correction (Chen et al., 2017), and calculated Cohen’s d for each second-wave MHG vs. 12-step.

Last, to identify which barrier subscales were most predictive of MHG involvement over 12 months (Aim 3), we examined lagged associations, in which barrier subscale scores at baseline predicted MHG involvement at 6-months, and barrier subscale scores at 6-month predicted MHG involvement at 12-month, using Generalized Estimating Equations (Liang & Liu, 1991). The models used a Gaussian (normal) distribution and an identity link function; produced cluster-robust standard errors; and had an unstructured within-group correlation structure. We standardized all continuous independent variables and the dependent variable (mean = 0, SD = 1) to yield standardized regression coefficients. We entered categorical variables as unstandardized dummy variables.

Unadjusted models included only the lagged barrier subscale scores as predictors. Adjusted models included covariates if they were significantly associated with MHG involvement at 6 or 12 months in bivariate analyses; they controlled for sociodemographics (age as a continuous variable, race, ethnicity, marital status, educational attainment, employment status, income, and urbanicity), clinical variables (lifetime drug problems, current alcohol recovery goal, last alcohol use, lifetime AUD severity, and lifetime psychiatric severity), and MHG characteristics (primary MHG and mutual-help attendance mode in the past 30 days).

RESULTS

Sample characteristics by primary group at baseline

Table 1 shows baseline characteristics by primary group. Participants who selected WFS, LifeRing, or SMART (vs. 12-step) as their primary group were older and more likely to be male (except for those selecting WFS, who were more likely to be female) and White. Participants who selected WFS and LifeRing (vs. 12-step) were also more likely to be married/partnered, have a college degree, and have higher incomes, and were less likely to report lifetime drug use problems. Finally, past-30 attendance was higher among 12-step attendees (vs. second-wave MHGs).

REASONS scale analyses and overall item and subscale means

Table 2 shows the results of preliminary analyses of the REASONS scale along with item and subscale means to describe the most common barriers to continued attendance. We identified 2 underlying factors with eigenvalues greater than 1. Factor One contained 6 items from the original motivation/no perceived need, dislike of meeting attendees, and dislike of meeting content domains. For increased interpretability, we divided these items into two subscales, one encompassing motivation and perceived need (3 items) and the other encompassing meeting attendees and content (3 items). Factor Two contained 5 items from the original social anxiety and psychiatric concerns domains, treated as one subscale. Supporting use of these three subscales, analyses of the original REASONS measure identified significant correlations between the dislike of meeting content and dislike of meeting attendees subscales, and between the social anxiety and psychiatric concerns subscales, but no significant associations between the motivation/no perceived need subscale and any other subscales (Kelly et al., 2010). The adapted REASONS scale baseline alpha was 0.87, and 6-month alpha was 0.90.

The social anxiety and psychiatric concerns subscale had the highest mean of the three subscales (mean=1.80, SD=0.85), and the subscale’s most highly endorsed item was “I don’t like crowds or large groups” (mean=2.08, SD=1.26). The dislike of meeting attendees and content subscale had the second highest mean (mean=1.74, SD=0.85), and the subscale’s most highly endorsed item was “I didn’t like hearing the same stories over and over again” (mean=1.86, SD=1.10). Low motivation or perceived need had the lowest mean (mean=1.69, SD=0.90) and the subscale’s most highly endorsed item was “I thought I could change and do it myself” (mean=1.90, SD=1.23).

Differences in barriers subscales across primary group

The baseline barrier subscale means by primary group are presented in Table 3. All barrier subscale means were lower among second-wave MHG (vs. 12-step) primary group attendees. The social anxiety and psychiatric concerns subscale had, descriptively, the highest mean among all MHGs except LifeRing, where it had the lowest mean. Among LifeRing attendees, the dislike of meeting attendees and content and the low motivation or perceived need subscales were equally the highest. Cohen’s d values indicated moderate to large differences across groups, particularly for the social anxiety and psychiatric concerns subscale (d = 0.56–0.79). Effect sizes for the dislike/content subscale ranged from 0.50–0.75, while those for low motivation ranged from 0.27–0.52.

Table 3. Mean Barrier Subscales Scores by Primary Group at Baseline.

	12-step	WFS	d vs. 12-step	LifeRing	d vs. 12-step	SMART	d vs. 12-step
Barrier subscale (range: 1–5)	Mean (SD)	Mean (SD)		Mean (SD)		Mean (SD)
Low motivation/perceived need	1.99 (0.96)	1.59 (0.84) **	0.44	1.73 (0.97)	0.27	1.53 (0.80) ***	0.52
Dislike of meeting attendees and content	2.21 (1.02)	1.55 (0.71) ***	0.75	1.73 (0.90) ***	0.50	1.58 (0.66) ***	0.73
Social anxiety and psychiatric concerns	2.29 (0.98)	1.78 (0.84) ***	0.56	1.61 (0.72) ***	0.79	1.68 (0.76) ***	0.70

Note: p-values use 12-step as the referent group

^*p-value<0.05

^**p-value<0.01

^***p-value<0.001; Cohen’s d reflects standardized mean differences in barrier subscale scores between each second-wave MHG group and the 12-step group.

Barrier subscales as time-lagged predictors of MHG involvement

Longitudinal associations between lagged barrier subscale scores and MHG involvement at 6- and 12-month follow-ups are presented in Table 4. In the unadjusted model, higher scores on the social anxiety and psychiatric concerns subscale predicted lower MHG involvement (β=−0.10 (−0.18, −0.09), p<0.05), with no other subscales predicting involvement. This association remained significant in the adjusted model (β=−0.09 (−0.18, −0.01), p<0.05). Higher MHG involvement was also associated with having an income between $40,001 and $80,000 per year, higher lifetime AUD severity, using alcohol 1+ (vs. less than 1) years ago, and attending meetings both in-person and online (vs. in-person only). Lower MHG involvement was associated with attending MHGs online only (vs. in-person only).

Table 4. Longitudinal Associations between Lagged Barrier Subscale Scores and Primary Group Involvement at 6 and 12 months.

	Unadjusted			Adjusted
	Beta	CI	p-value	Beta	CI	p-value
Barrier subscales
Not ready	-0.05	(−0.12, 0.01)	ns	0.03	(−0.04, 0.10)	ns
Meeting issues	0.04	(−0.04, 0.11)	ns	0.02	(−0.06, 0.10)	ns
Social and psychiatric concerns	-0.10	(−0.18, −0.09)	*	-0.09	(−0.18, −0.01)	*
Sociodemographics
Age				0.06	(−0.04, 0.17)	ns
Gender
Female (vs. male)				0.10	(−0.10, 0.30)	ns
Race and ethnicity
Non-Hispanic White				ref	ref
Non-Hispanic Black/African American				0.04	(−0.37, 0.45)	ns
Latinx/Hispanic				−0.04	(−0.36, 0.28)	ns
Other				0.32	(−0.11, 0.74)	ns
Marital Status
Single				ref
Separated/widowed/divorced				0.02	(−0.22, 0.27)	ns
Married/partnered				0.03	(−0.22, 0.28)	ns
Educational Attainment
Less than college degree				ref	ref
College degree (AA/BA)				−0.12	(−0.34, 0.08)	ns
Post-graduate training or degree				−0.15	(−0.41, 0.10)	ns
Income
40,000 or less				ref	ref
40,001 to 80,000				0.27	(0.01, 0.15)	*
80,001 to 120,000				0.13	(−0.14, 0.39)	ns
More than 120,000				0.11	(−0.19, 0.41)	ns
Urbanicity
Urban/suburban (vs. rural)				−0.05	(−0.27, 0.17)	ns
Employment Status
Employed full/part-time (vs. unemployed)				−0.05	(−0.24, 0.14)	ns
Clinical characteristics
Lifetime AUD severity				0.19	(0.10, 0.27)	***
Lifetime 2 drug use problems
Yes (vs. no)				0.06	(−0.12, 0.24)	ns
Lifetime psychiatric severity				−0.08	(−0.17, 0.01)	ns
Current alcohol recovery goal
Total lifetime abstinence (vs. other)				0.18	(−0.01, 0.37)	ns
Last alcohol use
1 or more years (vs. less than 1 year) ago				0.34	(0.14, 0.54)	**
Mutual Help Group characteristics
Primary Help Group Choice
12-step				ref	ref
WFS				−0.08	(−0.30, 0.15)	ns
LifeRing				−0.11	(−0.32, 0.11)	ns
SMART				−0.15	(−0.38, 0.07)	ns
Other				−0.19	(−0.61, 0.23)	ns
No past 30-day attendance				−0.003	(−0.66, 0.65)	ns
Mutual-help attendance mode past 30 days
In-person only				ref	ref
Online only				−0.24	(−0.45, −0.03)	*
Both in-person and online				0.22	(0.03, 0.42)	*
No attendance past 30 days				−0.24	(−0.49, 0.02)	ns

Note:

^*p-value<0.05

^**p-value<0.01

^***p-value<0.001; ns = not significant; beta coefficients are standardized for continuous variables only.

Supplementary Tables

Supplementary Tables 1 and 2 show results of analyses to empirically support the selection of MHG involvement vs. attendance as the outcome. MHG attendance and involvement showed weak to moderate correlations (r=0.26–0.32), and only involvement was significantly associated with alcohol use outcomes in both bivariate and multivariate models.

DISCUSSION

This study aimed to describe the most common barriers to continued MHG attendance among participants of diverse MHGs; to examine variation in barriers across primary MHG; and to identify the barrier domains most predictive of MHG involvement. Overall, social anxiety and psychiatric concerns emerged as the most highly endorsed barrier subscale in the overall sample and for all MHGs except LifeRing. Also, the most highly endorsed individual barrier item reflected social anxiety (“I don’t like crowds or large groups”). Finally, higher time-lagged social anxiety and psychiatric concerns subscale scores predicted lower MHG involvement at 6- and 12-month follow-ups, and this was the only barrier domain predicting subsequent MHG involvement. The salience of social anxiety and psychiatric barriers suggests that efforts to improve the social experiences of MHG attendees could improve MHG attendance, involvement, and ultimately, recovery outcomes. For example, MHGs could encourage meetings of smaller sizes to alleviate anxiety about large groups and make special efforts to help newcomers feel welcome and connect socially. MHGs could also dedicate portions of meetings to sharing experiences of psychiatric problems without judgement or criticism. Additionally, providers could recommend MHGs supportive of people with co-occurring disorders to clients with concerns about sharing psychiatric experiences. Indeed, there are a variety of effective MHGs for people with co-occurring substance use and mental health disorders that providers could ensure their clients are aware of and encouraged to attend (Bergman et al., 2014; Magura, 2008; Vogel et al., 1998). Encouragingly, previous work by Thevos et al found that participation in 12-step groups was associated with improved social support among people with AUD and co-occurring social phobia (Thevos et al., 2001), underscoring the potential value of fostering positive social experiences for MHG attendees.

Our finding that higher scores on the social anxiety and psychiatric concerns subscale predicted lower MHG involvement contrasts with findings from previous work showing people with co-morbid mental health disorders benefit from MHG participation as much (Kelly et al., 2003) or more than people with less severe or no co-morbidity (Bergman et al., 2014; Laudet et al., 2003; Timko et al., 2013). Indeed, a meta-analysis by Tonigan et al found that AA attendance by dually-diagnosed individuals was significantly and positively associated with increased alcohol abstinence (Tonigan et al., 1996). A possible explanation for this discrepancy is that the social anxiety and psychiatric concerns subscale did not measure psychopathology per se, as these previous studies did. Further, our measure of lifetime psychiatric severity did not differ across MHGs, suggesting psychopathology was not as relevant in our sample and to our findings.

The original REASONS measure had social anxiety barriers and psychiatric barriers as separate subscales and had 24 items in contrast to our 11-item version, complicating direct comparisons of findings across versions. Nonetheless, in the REASONS development study (Kelly et al., 2010), the item, “I thought I could change and do it myself” had the highest item mean, and the motivation/no perceived need subscale had the highest subscale mean. Differences in findings across studies might be explained by the different sample demographics (e.g., Kelly examined male veterans newly entering SUD treatment with heterogeneous SUD diagnoses, whereas the current study examined majority-female community members with primarily AUD diagnoses). Other differences are that Kelly examined barriers to attending 12-step meetings among individuals who may have ceased attendance, whereas we examined barriers to continued attendance among individuals currently attending diverse groups. Even with these differences, both the “I thought I could change and do it myself” item and social anxiety items generally were highly rated in both studies, suggesting the relevance of perceived need for recovery support and concerns about social interactions as barriers across different demographic groups of people struggling with substance use.

No studies have previously identified barriers to attendance among second-wave MHGs or examined differences in barriers between second-wave MHGs and 12-step MHGs. We found that all second-wave MHGs were lower on all barrier subscale domains compared to 12-step MHGs. Second-wave MHG attendees in this sample were older than 12-step attendees, and older age may be associated with greater motivation to obtain help for an AUD (Oslin et al., 2002). Another possible explanation is that second-wave MHGs generally aim to be highly supportive and welcoming environments. For example, they eschew religious content to provide a space to those deterred by the spiritual elements of AA (Schmidt, 1996; Tonigan et al., 2002) and, in the context of WFS, support the personal empowerment of women and encourage positive thinking and self-esteem (Fenner & Gifford, 2012). A final possible explanation is that people attending second-wave MHGs experience lower attendance barriers because, having tried12-step groups, they self-selected to an alternative with (for them) lower barriers to attendance. Similarly, the limited availability of second-wave MHGs relative to the ubiquity of 12-step groups may mean that second-wave attendees intentionally seek out these groups and are generally motivated to participate and thus perceive lower barriers, whereas 12-step groups may include a heterogenous group of individuals, with some drawn to the accessibility of meetings yet less motivated by the model itself.

This study has several limitations. First, we cannot guarantee the representativeness of our sample given that the MHGs under study do not manage complete lists of attendees, rendering it impossible to calculate response rates. Similarly, 12-step organizations do not directly engage in research, precluding a recruitment approach comparable to that for second-wave MHGs. Relatedly, as we used online surveys, participants with reliable internet access and comfort with online surveys may be over-represented in our sample. Still, all demographic comparisons of the current sample—stratified by primary group—to samples of AA, WFS, LifeRing, and SMART attendees collected for internal surveys have shown only minor differences except in the case of gender, with women substantially over-represented in the PAL Study 2021 Cohort 12-step sample, vs. AA’s internal survey data (Zemore, 2024). Also, while we based our barriers to attendance measure on the validated REASONS measure and applied appropriate statistical techniques to examine its underlying structure, we did not make comparisons to the original measure or test other psychometric properties of the adapted version. Additionally, we did not control for the length of time in recovery or length of time attending MHGs, both of which could impact barriers to attendance and involvement, and the past 30-day attendance measure may have missed attendance outside of that timeframe at the follow-up interviews. Finally, we collected our sample during the second year of the COVID-19 pandemic, which may have influenced the social anxiety and psychiatric concerns measure given that the pandemic was a time of health concern for social interactions. This is especially relevant to the “I don’t like crowds or large groups” item, although less relevant to the 4 other items. We also note that baseline data collection began nearly one year after COVID-19 vaccines became available, and that we did not observe any significant changes in social anxiety and psychiatric concerns subscale scores between baseline and 6 months (data not shown). Nonetheless, the inability to disentangle any COVID-19 effects on participant responses highlights the need to replicate our findings.

Despite limitations, our findings are useful for guiding future work and AUD interventions. Our observation that barriers to attendance are lower among second-wave MHGs is encouraging and supports provider referral to these alternatives. Additionally, findings that higher social anxiety and psychiatric barriers predicts lower subsequent MHG involvement suggest that addressing these barriers could improve engagement with MHGs and support alcohol recovery. Providers and MHGs should consider making explicit that it is acceptable and even wise for attendees to try different meetings to find what makes them feel most comfortable and safest, thereby maximizing the likelihood of successful support for their recovery goals.