Mode of Mutual-Help Group Attendance: Predictors and Outcomes in a US National Longitudinal Survey of Adults with Lifetime Alcohol Use Disorder

Abstract

Introduction:

Although attending substance use-focused mutual-help meetings online may reduce attendance barriers, associations of attendance mode with group participation and outcomes are unknown. Using longitudinal data from mutual-help group attendees, this study, after identifying differences in baseline characteristics by attendance mode, examined associations of attendance mode with mutual-help participation (number of meetings attended, involvement) and outcomes (alcohol abstinence, heavy drinking, alcohol problems).

Methods:

The Peer Alternatives for Addiction Study 2021 Cohort sampled attendees of 12-step groups (e.g., Alcoholics Anonymous), Women for Sobriety, LifeRing Secular Recovery, and/or SMART Recovery in-person and/or online within 30 days before baseline. The baseline sample, recruited in fall 2021, was 531 adults with lifetime alcohol use disorder, followed at 6 (88%) and 12 months (85%). Differences in baseline characteristics by attendance mode were tested using Chi-squares and ANOVAs. GEE models examined associations of attendance mode, time, and their interactions with mutual-help group participation and alcohol outcomes. The in-person only mode was compared to the online-only, and to the in-person plus online, modes.

Results:

At baseline, 53.7% of participants had attended only online meetings in the past 30 days, 33.7% had attended both in-person and online meetings, and 12.6% had attended only in-person meetings. Online meeting attendees were less likely to endorse lifetime abstinence as an alcohol recovery goal than in-person-only meeting attendees. In adjusted models (including for recovery goal), those attending online meetings only, or both online and in-person meetings, attended a greater number of meetings compared to those attending only in-person meetings. However, online-only attendance was associated with less involvement than in-person-only attendance. In adjusted models, compared to baseline, involvement increased and outcomes improved at follow-ups. Adjusted models examining alcohol outcomes found that no attendance at mutual-help groups at follow-ups was associated with more heavy drinking compared to in-person-only attendance.

Conclusions:

Findings inform efforts to ascertain benefits of mutual-help group participation by suggesting that online attendance is associated with attending more meetings, less involvement, and lower endorsement of abstinence as a recovery goal, and is comparable to in-person attendance on alcohol outcomes. In-person attendance may be more beneficial for less heavy drinking than terminating attendance.

1. Introduction

Attending substance use-focused mutual-help meetings online holds promise to increase access and reduce barriers to group participation. Indeed, in response to social distancing necessitated by the COVID-19 pandemic, meetings using online video platforms were the only way to attend substance use-focused mutual-help organizations (Bergman et al., 2021). With the easing of pandemic-related restrictions and resumption of in-person meetings, mutual-help programs have continued to accommodate online attendance in both all-virtual meetings and using hybrid formats with a mix of in-person and remote attendees. Given the promise of increased access to mutual-help groups with the availability of online meetings, people seeking help for substance use, treatment providers, and researchers want more information on characteristics of online attendees, and on benefits and drawbacks associated with mode of mutual-help group attendance.

Emerging literature has begun to examine the characteristics of people attending mutual-help groups for substance use online. In a previous study using data from the US Peer Alternatives for Addiction (PAL) Study 2015 Cohort (Zemore et al., 2017, 2018), Timko et al. (2022) examined online meeting attendance among individuals who had attended an in-person meeting of a 12-step group, Women for Sobriety (WFS), LifeRing Secular Recovery (LifeRing), and/or SMART Recovery (SMART) for an alcohol problem of their own in the 30 days prior to baseline. At baseline, 62% (of 402 participants) had attended an online mutual-help group meeting in their lifetime, and 36% had done so in the past 30 days. Online meeting attendance in the past 30 days was more likely among participants with more recent alcohol and drug use, and those with less abstinence self-efficacy. In a US national survey of adults with a resolved alcohol problem (n=1,487), 15% reported lifetime use of Digital Recovery Support Services (D-RSS, defined as online technology-based modalities, e.g., online video recovery support meetings, discussion boards, chat rooms, social networking sites) (Gilbert et al., 2022). Compared to lifetime D-RSS nonusers, D-RSS users had larger proportions of persons with severe lifetime alcohol use disorder (AUD), who had obtained treatment, and had been in recovery less than five years.

A US national survey of 2,152 Narcotics Anonymous (NA) long-term attendees examined participants’ views of online and in-person meeting attendance (Galanter et al., 2022). Most participants (65%) agreed that online meetings were as good as or better than in-person meetings for maintaining their own abstinence. However, only 42% agreed that online meetings were good as or better than in-person meetings for maintaining a newcomer’s abstinence. Number of meetings increased with the switch from in-person to online meetings set off by the pandemic (Galanter et al., 2022).

A study in Australia of online SMART Recovery groups (Beck et al., 2023b) found that over 94% of participants (n=1,414) felt engaged (welcomed, supported, and understood, with opportunities to contribute), thought groups were well-facilitated and contributed to their recovery, and planned to continue to attend SMART Recovery online. In an evaluation of online mutual-help groups by SMART Recovery Australia, of 345 participants who had attended both in-person and online meetings, online meetings were rated as better or much better than in-person meetings by 52%, and as about the same by 37% (Beck et al., 2023a).

Distinct from these findings of positive views of online meetings, a study in Poland of 225 Alcoholics Anonymous (AA) attendees found that although the frequency of meeting attendance and contact with sponsors increased after meetings switched (due to the pandemic) from in-person to online, satisfaction from participation decreased (Hoffmann & Dudkiewicz, 2021). After the switch, participants felt less of a sense of community, support, friendship, and sharing. They also felt less of a drive toward personal development, learning how to listen to others and accept their honesty and candor, and acquiring a realistic self-image. Two-thirds of respondents preferred mainly in-person AA meetings complemented by occasional meetings online (Hoffmann & Dudkiewicz, 2021).

Similarly, a survey of 97 12-step community members from IntheRooms.com (a recovery social networking site) found that members perceived more challenges than advantages to be associated with online compared to in-person meetings (Barrett & Murphy, 2021). Moreover, in-person 12-step meetings were associated with perceptions of greater meeting effectiveness and support and recovery network quality. Findings suggested that online meetings are a valuable supplemental tool but should not replace in-person meetings (Barrett & Murphy, 2021). These findings agree with those from a qualitative focus group study of 17 attendees of 12-step programs from six US states (Senreich et al., 2022). Based on their experiences attending online and in-person meetings, respondents praised the accessibility and convenience of online meetings, and the ability to interact with members from different geographic locations. However, they noted the loss of intensive social support found at in-person meetings and technological challenges of online meetings (Senreich et al., 2022).

While the cited studies shed some light on characteristics and experiences of people who use online recovery support services, very little empirical research has examined associations between mode of mutual-help group attendance and mutual-help group participation and substance use outcomes. In the PAL Study 2015 Cohort, longitudinal analyses examining concurrent attendance mode and abstinence outcomes found that, compared to participants who attended in-person meetings only, those attending both in-person and online meetings were less likely to be abstinent at baseline. However, attendance mode was unrelated to abstinence at the 12-month follow-up (Timko et al., 2022).

The present study used longitudinal data from the PAL Study 2021 Cohort to fill key gaps in the emerging literature on corollaries of mode of mutual-help group attendance. Like the PAL Study 2015 Cohort, the 2021 Cohort was designed to examine the comparative effectiveness of abstinence-focused mutual-help group alternatives to 12-step groups prevalent in the US (i.e., WFS, LifeRing, SMART) for supporting recovery from alcohol problems (Zemore et al., 2017, 2018). In the PAL Study 2021 Cohort, all participants had attended at least one in-person or online mutual-help group meeting in the 30 days prior to baseline (in contrast to the PAL Study 2015 Cohort, in which in-person attendance was required).

The present study first examined associations between mode of attendance (in-person only, online only, or both in person and online) and participants’ demographic and clinical characteristics at baseline. This first step identified characteristics of participants that varied across attendance mode and needed to be controlled in the study’s analytic models. Then, using all surveys, we explored how mode of attendance related to mutual-help group participation, that is, meeting attendance and program involvement. Mutual-help involvement is comprised of engagement in program practices or activities, and is important to examine because it may be sustained over time even when meeting attendance declines, and is a key aspect of mutual-help group participation that is associated with better substance use outcomes (Costello et al., 2021; Timko et al., 2013). Last, and again using all surveys, we explored how mode of attendance related to alcohol outcomes, including alcohol abstinence, heavy drinking, and alcohol problems. Because not all participants at baseline continued mutual-help group attendance prior to follow-ups, examinations of mode with outcomes included participants who were no longer attending, as a separate group.

These analyses significantly extend prior work by examining associations of mode of mutual-help group attendance with attendees’ characteristics and with their group participation and alcohol outcomes. Although previous studies found that meeting attendance increased after the switch from in-person to online meetings (Galanter et al., 2022; Hoffmann & Dudkiewicz, 2021), studies did not directly compare attendance by mode. In addition, previous studies have not examined mutual-help group mode’s association with program involvement. Results from the present study will help substance use treatment providers and courts make informed referrals to mutual-help groups, and help people seeking peer support to make choices about their recovery pathways.

2. Method

2.1. Sample and procedure

Recruited respondents completed baseline surveys used in this study (n=531) from October through December 2021. The leadership of WFS, LifeRing, and SMART collaborated on recruitment and publicized the survey via email and social media (e.g., their websites) and at meetings. In addition, the study recruited 12-step group attendees via collaborations with IntheRooms and Faces and Voices of Recovery (which advocates for public policies and funding that support substance use disorder recovery). IntheRooms promoted the study via social media and emails to their community. Faces and Voices of Recovery endorsed the study to recovery community organization directors at meetings and by email. Individuals interested in participating accessed the survey on the PAL Study website and completed the online screen to determine their eligibility. All participants were required to be at least 18 years old; a US resident; report attending at least one in-person or online 12-step, WFS, LifeRing, or SMART meeting for an alcohol and/or drug problem of their own in the past 30 days; report a lifetime AUD; and not have participated in the PAL Study 2015 Cohort. Lifetime AUD was determined using Composite International Diagnostic Interview (CIDI) items that are in the alcohol section (WHO, 1993). Eighteen items addressed the 11 criteria for a DSM-5 AUD diagnosis (APA, 2013); individuals endorsing 2+ criteria are considered to have a lifetime AUD disorder. Only participants who were screened and met these criteria had access to the baseline survey.

Of 5,075 screens completed, 798 were disqualified because the participant was <18 years of age (n=8), was not a US resident (n=11), did not have a lifetime AUD (n=145), had no or non-qualifying mutual-help group attendance (attendance at a group that was not one of the study’s included groups; n=301), was in the PAL Study 2015 Cohort (n=59), and/or would have exceeded study recruitment quotas for each mutual-help group and/or recruitment source (n=274) (see Supplementary Table 1). Recruitment quotas called for subgroups of about equal size of individuals attending 12-step groups, WFS, LifeRing, and SMART. Of 4,277 eligible participants with initiated surveys, 3,968 completed surveys (92.8%). The study systematically reviewed completed surveys at each timepoint using a multi-pronged strategy to eliminate surveys identified as fraudulent or containing low-quality data. Of 3,968 completed baseline surveys, 3,437 were deemed fraudulent or low-quality (e.g., identified as having an invalid and/or non-residential address). Follow-up surveys were conducted 6 and 12 months post-baseline (88% [n=464] and 85% [n=449] response rate, respectively, of retained baseline surveys). Participants received Amazon gift cards for completing surveys ($35 at baseline and $40 for each follow-up). Participants lost to follow-up at 6 months were more likely to be female (OR=1.90, p<0.05) and less likely to report lifetime drug problems (OR= 0.55, p<0.05) or SMART compared to 12-step as their primary group at baseline (OR=0.47, p<0.05). Participants lost to follow-up at 12 months were more likely to be female (OR=2.36, p<0.01) and less likely to report LifeRing or SMART compared to 12-step as their primary group at baseline (OR=0.45, p<0.05 and OR=0.43, p<0.05, respectively).

Measures

2.2.1. Mutual-help group participation

At baseline and follow-ups, participants reported the number of in-person and online 12-step, WFS, LifeRing, and SMART meetings they had attended in the past 30 days for an alcohol or drug problem of their own. All respondents had past-30-day attendance at baseline, but no attendance in the past 30 days was reported by 80 participants (17%) at 6 months, and by 82 (18%) participants at 12 months. Responses were used to code attendance mode aggregated across all groups attended at each timepoint: in-person only, online only, or both modes. In addition, numbers of meetings attended (whether in person or online) were summed across all groups (12-step, WFS, LifeRing, and SMART) to yield total meeting attendance. Given the highly non-normal distribution, and literature reporting the mean number of 12-step meetings attended per month to be 8 to 9 (Humphreys et al., 2014; Kaskutas et al., 2005; Witbrodt et al., 2014), or meetings attended per week to be 2-4 (Donovan et al., 2013), meeting count was dichotomized into 8+ (i.e., at least 2 meetings per week) vs. <8 meetings (i.e., fewer than 2 meetings per week) in the past 30 days.

Mutual-help group involvement in the past 30 days was assessed at each timepoint using 8 yes/no items. Four items were adapted from a standard scale of 12-step group involvement, the AA Affiliation Scale (Humphreys et al., 1998). These items were strongly associated with substance use outcomes (Zemore et al., 2013, 2018) and are worded to be appropriate for all groups under study, not just AA groups. They inquired whether participants 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 in the past 30 days; and did volunteer work or “service” at a meeting in the past 30 days. An additional 4 items were developed for the PAL Study 2021 Cohort to capture mutual-help group involvement across in-person and online modes more fully. Items were generated from two focus groups with individuals currently attending the mutual-help groups under study. They inquired whether, in the past 30 days, participants shared openly during a meeting; met in person outside of meetings with others who attended the same meetings; connected one-on-one via text, email, phone, or video chat with others who attended the same meetings; and spent time socializing with other meeting attendees just before and/or after meetings. A summary measure was created by averaging across all 8 items, with yes=1 and no=0 (baseline alpha=0.73).

2.2.2. Alcohol outcomes

Alcohol outcomes were assessed at all timepoints. Alcohol abstinence was indicated by self-reported abstinence from alcohol during the past 6 months. Heavy drinking was measured as the count (0-30) of days during the past 30 that participants reported at least 4 (for women) or 5 (for men) drinks containing alcohol in a single day. Participants were instructed to count any combination of alcoholic beverages, such as beer, wine, spirits, and cocktails, and to give their best guess even if unsure of the exact number. Alcohol problems were assessed using 5 items from the Short Index of Alcohol Problems (SIP) selected on the basis of factor analysis, domain coverage, and the benefit of brevity to reduce respondent burden (Feinn et al., 2003), i.e., have not eaten properly, failed to do what is expected of me because of my drinking, taken foolish risks when I have been drinking, been hurt by my drinking, and drinking has gotten in the way of my growth as a person. Items referred to the past 12 months at baseline (alpha=0.75) and past 6 months at follow-ups (alphas=0.73 and 0.81, respectively). Responses were used to create an indicator of any problems vs. none.

2.2.3. Covariates

Covariates were selected based on their relationship to attendance characteristics and alcohol outcomes and included demographic, clinical, and recovery-related factors. Demographic characteristics, assessed at baseline, included gender (see notes on Table 1), age, race/ethnicity (Non-Hispanic White, Non-Hispanic Black/African-American, Other), marital status (single, never married; separated, divorced, or widowed; married or partnered), education (less than college degree, college degree, post-college degree), employment status (unemployed, employed), annual household income (≤40,000 US dollars, $40,001-$80,000, $80,001-$120,000, >$120,000), and urbanicity (urban, suburban, rural).

Table 1.

Baseline Characteristics by Mutual-Help Group Attendance Mode (n=531).

	Total Sample (N=531)		Past-30-Day Attendance Mode
			In-person Only (N=67)		Online Only (N=285)		Both (N=179)
	n	%	n	%	n	%	n	%	χ2	p
Gender									2.16	0.339
Male	177	33.71	27	40.91	95	33.81	55	30.90
Female	348	66.29	39	59.09	186	66.19	123	69.10
Race/Ethnicity									9.24	0.055
Non-Hispanic White	448	84.37	61	91.04	244	85.61	143	79.89
Non-Hispanic Black/African American	33	6.21	3	4.48	12	4.21	18	10.06
Other	50	9.42	3	4.48	29	10.18	18	10.06
Marital Status									18.48	0.001
Single, never married	122	22.98	10	14.93	56	19.65	56	31.28
Separated/divorced/widowed	126	23.73	21	31.34	58	20.35	47	26.26
Married/living with partner	283	53.30	36	53.73	171	60.00	76	42.46
Education									15.32	0.004
Less than college degree	142	26.74	20	29.85	58	20.35	64	35.75
College degree	267	50.28	34	50.75	150	52.63	83	46.37
Post-graduate training or degree	122	22.98	13	19.40	77	27.02	32	17.88
Employment Status									1.20	0.548
Unemployed/not working	201	37.85	26	38.81	113	39.65	62	34.64
Employed full/part-time	330	62.15	41	61.19	172	60.35	117	65.36
Income (US $)									17.42	0.008
40,000 or less	161	30.43	17	25.76	76	26.76	68	37.99
40,001 to 80,000	135	25.52	19	28.79	69	24.30	47	26.26
80,001 to 120,000	138	26.09	11	16.67	86	30.28	41	22.91
More than 120,000	95	17.96	19	28.79	53	18.66	23	12.85
Urbanicity									3.84	0.429
Urban	177	33.33	17	25.37	96	33.68	64	35.75
Suburban	253	47.65	37	55.22	130	45.61	86	48.04
Rural town/rural area	101	19.02	13	19.40	59	20.70	29	16.20
Drug Use Problems (lifetime)									8.67	0.013
No	248	46.79	30	44.78	149	52.46	69	38.55
Yes	282	53.21	37	55.22	135	47.54	110	61.45
Current Alcohol Recovery Goal									7.67	0.022
Not lifetime abstinence	256	48.30	22	32.84	147	51.58	87	48.88
Total lifetime abstinence	274	51.70	45	67.16	138	48.42	91	51.12
Primary Mutual-help Group									89.49	0.000
12-step	105	19.77	9	13.43	28	9.82	68	37.99
LifeRing	136	25.61	13	19.40	98	34.39	25	13.97
SMART	159	29.94	35	52.24	70	24.56	54	30.17
WFS	131	24.67	10	14.93	89	31.23	32	17.88
	M	SE	M	SE	M	SE	M	SE	F	p
Age	47.83	0.59	49.46	1.40	49.12	0.85	45.17	0.96	5.29	0.005
Lifetime AUD Criteria	9.70	0.07	9.55	0.23	9.49	0.10	10.09	0.10	7.98	0.000

Notes. Valid column percentages are presented. Missing data were minimal and primarily due to non-response. However, 6 individuals who reported their gender as other than male/female were recoded to missing to better examine differences in mode of attendance between males and females. Differences between modes of attendance were examined with Chi-square and ANOVA tests.

Lifetime AUD severity was assessed at baseline using the 18 items from the CIDI that addressed the 11 criteria for a DSM-5 AUD diagnosis. As in previous studies (Cherpitel et al., 2015; Tsutumi et al., 2020; Zemore et al., 2017, 2018), AUD severity was coded as a count of criteria endorsed, ranging from 2 to 11. Baseline surveys also assessed lifetime drug use problems using 2 yes/no items asking whether there were “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 “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 use problems, and those who endorsed one or no items were coded as not having these problems. The two items are used in the DSM-5 criteria for drug use disorders and have a high prevalence of endorsement across different drugs (see Borges et al., 2015; Saha et al., 2012).

Alcohol recovery goal was also assessed at each timepoint using a single item (Hall et al., 1991) asking respondents to select the recovery goal among 5 presented that was most true for them, ranging from total lifetime abstinence to controlled use. Abstinence goals have been found to be related to lower risk of relapse (Hall et al., 1991) and versions are often used in clinical assessments. We created a dichotomous measure in which responses were coded as total lifetime abstinence if participants endorsed, “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.” All other responses (e.g., “I want to use alcohol in a controlled manner”) were coded as another goal.

At each timepoint, the survey collected participants’ primary mutual-help group. For participants who reported attending only one of the four groups (i.e., 12-step, WFS, LifeRIng, SMART) in the past 30 days, that group was coded as their primary group. Participants who attended more than one of these groups in the past 30 days self-identified their primary group.

2.3. Analysis

To begin, we examined differences in baseline sample characteristics by past-30-day attendance mode at baseline using Chi-square tests for categorical variables and ANOVAs for continuous measures. These analyses were used to describe characteristics of participants by attendance mode, and to identify participant characteristics that should be controlled in subsequent analyses. To examine associations between attendance mode and both mutual-help group participation (attendance, involvement) and alcohol outcomes across surveys, the study used a generalized estimating equation (GEE) approach. We used this approach, which is commonly used in longitudinal data analysis, because models can account for within-participant non-independence of observations (i.e., nesting or clustering) across multiple waves of data collection. Specifically, models first examined the independent associations of attendance mode (in-person only as the reference group, online only, or both in-person and online) and time (baseline [reference], 6-month, and 12-month surveys) with mutual-help group participation among those who attended mutual-help groups in the past 30 days (n=531 at baseline, n=384 at 6-month follow-up, and n=367 at 12-month follow-up). Models then examined associations of mode and time with alcohol outcomes, including those with no mutual-help group participation in the past 30 days at follow-ups (n=531 at baseline, n=464 at 6 months, and n=449 at 12 months), with attendance mode and outcomes measured concurrently. Following the standard GEE approach, we then added interactions between attendance mode and time, in order to test whether time moderated effects of attendance mode. Time was modeled as a categorical rather than a continuous variable to more precisely identify when changes occurred. When a significant interaction was found, predictive margins were plotted to depict its nature. Models were first run without covariates (unadjusted models) and then with covariates (adjusted) to determine whether findings remained after adjusting for potential confounders, which are noted in Tables 2 and 4.

Table 2.

Associations of Attendance Mode and Time with Mutual-help Group Participation Among Attendees of Meetings in the Past 30 Days.

	Meeting Attendance (8+ vs Fewer)						Involvement
	Unadjusted Model			Adjusted Model			Unadjusted Model			Adjusted Model
	OR	95% CI	p	OR	95% CI	p	Coef	95% CI	p	Coef	95% CI	p
Timepoint
Baseline (Ref) (n=541)
6-mo (n=384)	1.06	[0.87, 1.29]	0.543	1.05	[0.83, 1.33]	0.691	0.04	[ 0.02, 0.06]	0.000	0.04	[ 0.02, 0.06]	0.000
12-mo (n=367)	1.08	[0.88, 1.32]	0.468	1.08	[0.86, 1.37]	0.505	0.03	[ 0.01, 0.05]	0.016	0.02	[−0.00, 0.05]	0.052
Wald test			0.740			0.800			0.000			0.000
Attendance Mode
In-person Only (Ref)
Online Only	1.11	[0.75, 1.64]	0.598	1.58	[1.02, 2.44]	0.043	−0.16	[−0.20, −0.11]	0.000	−0.13	[−0.18, −0.08]	0.000
Both	2.93	[1.95, 4.41]	0.000	4.39	[2.76, 7.00]	0.000	−0.03	[−0.08, 0.01]	0.142	−0.02	[−0.06, 0.02]	0.317
Wald test			0.000			0.000			0.000			0.000

Notes. Adjusted models add covariates of race/ethnicity, age, marital status, education, income, lifetime AUD severity, lifetime drug use problems, current alcohol recovery goal, and current primary mutual-help group. OR=Odds Ratio; CI=Confidence Interval; Coef=Coefficient.

Table 4.

Associations of Attendance Mode and Time with Alcohol Outcomes.

	Alcohol Abstinence in Past 6 Months			Heavy Drinking Days in Past 30 Days			Alcohol Problems in Past 12 Months/Since Last Survey
	OR	95% CI	p	IRR	95% CI	p	OR	95% CI	p
Unadjusted Models
Timepoint
Baseline (Ref) (n=541)
6-month (n=464)	1.49	[1.27, 1.75]	0.000	0.48	[0.36, 0.65]	0.000	0.40	[0.33, 0.48]	0.000
12-month (n=449)	1.66	[1.39,1.97]	0.000	0.54	[0.36, 0.65]	0.000	0.34	[0.28, 0.42]	0.000
Wald test			0.000			0.000			0.000
Attendance Mode
In-person Only (Ref)
Online Only	0.75	[0.50, 1.13]	0.172	1.17	[0.68, 2.00]	0.570	1.33	[0.90, 1.97]	0.158
Both	0.66	[0.43, 1.01]	0.058	1.10	[0.65, 1.87]	0.720	1.59	[1.07, 2.38]	0.022
None	0.54	[0.35, 0.84]	0.006	2.50	[1.32, 4.73]	0.005	1.39	[0.87, 2.23]	0.167
Wald test			0.021			0.005			0.100
Adjusted Models
Timepoint
Baseline (Ref)
6-month	1.72	[1.37, 2.17]	0.000	0.43	[0.29, 0.66]	0.000	0.29	[0.23, 0.38]	0.000
12-month	1.96	[1.54, 2.50]	0.000	0.40	[0.26, 0.64]	0.000	0.23	[0.18, 0.31]	0.000
Wald test			0.000			0.000			0.000
Attendance Mode
In-person Only (Ref)
Online Only	0.91	[0.54, 1.53]	0.731	1.03	[0.50, 2.13]	0.937	1.20	[0.74, 1.94]	0.458
Both	0.76	[0.44, 1.31]	0.325	0.95	[0.44, 2.04]	0.892	1.45	[0.91, 2.30]	0.118
None	0.57	[0.32, 1.01]	0.052	2.23	[1.01, 4.94]	0.048	1.28	[0.73, 2.27]	0.392
Wald test			0.070			0.016			0.362

Notes. Adjusted models add covariates of race/ethnicity, age, marital status, education, income, lifetime AUD severity, lifetime drug use problems, current alcohol recovery goal, and current primary mutual-help group. OR=Odds Ratio; CI=Confidence Interval; IRR=Incident Rate Ratio, i.e., the change in heavy drinking days in terms of a percentage increase or decrease, with the precise percentage determined by the amount the IRR is either above or below 1.

3. Results

3.1. Differences in sample characteristics by attendance mode at baseline

Table 1 displays participants’ baseline characteristics by mode of mutual-help group attendance at baseline. At baseline, for the full sample of 531 participants, past-30-day mutual-help group attendance was reported as in-person-only by 67 (12.6%) participants, online-only by 285 (53.7%), and both in-person and online by 179 (33.7%). Among demographic characteristics assessed, attendance mode was related to marital status, income, and age. Participants who attended both in-person and online meetings were less likely to be married or partnered, and more likely to be in the lowest income category, than participants who attended in-person only or online only; participants using both modes were also younger. Attendance mode was also related to education such that participants who attended online only were more likely to have a post-college degree. Regarding clinical factors, attendance mode was associated with lifetime drug use problems and lifetime AUD severity. Drug use problems and AUD severity were greatest among participants with both in-person and online attendance and least among those with online attendance only. Attendance mode was further related to current alcohol recovery goal and primary mutual-help group. Reporting a goal of total lifetime abstinence and SMART as the primary group was most likely among participants with in-person attendance only. Mode of attendance did not differ by participant gender, race/ethnicity, employment status, or urban/suburban/rural location.

3.2. Mutual-help group participation among meeting attendees

Table 2 displays results of unadjusted and adjusted GEE models testing associations of attendance mode and timepoint with participation (attendance and involvement) among respondents who attended mutual-help groups in the past 30 days at baseline and follow-up. In the unadjusted model examining attendance, attending both in-person and online meetings, but not attending online only (vs. in-person only), was associated with greater odds of having attended 8+ meetings in the past 30 days (OR=2.93, p<0.001). Timepoint was not associated with meeting attendance in the unadjusted models. In the adjusted models, attending both in-person and online meetings was associated with increased odds of 8+ meeting attendance (OR=4.39, p<0.001), and so was attending online only (OR=1.58, p=0.043). In the adjusted models, timepoint was not associated with meeting attendance.

Regarding involvement, the unadjusted model revealed significant associations for mode of attendance and time. Online-only attendance, but not both online and in-person attendance, compared to in-person-only attendance, was associated with less involvement (Coeff=−0.16, p<0.001). Regarding timepoint, compared to baseline, involvement increased at the 6-month (Coeff=0.04, p<0.001) and at the 12-month follow-up (Coeff=0.03, p=.016). In the adjusted model, the association of online only mode and of 6-month timepoint on involvement remained, but the association of the 12-month timepoint compared to baseline became only marginally significant (Coef=0.02, p=0.052).

There was a significant mode by time interaction with respect to attendance in the unadjusted model (Wald Test 2(4)=10.01, p=0.040; model coefficients not shown). Attending 8+ meetings for respondents attending both in-person and online increased compared to baseline at 6-month follow-up but became more similar to the two other groups at the 12-month follow-up, when the in-person-only group increased on attendance (Figure 1). Table 3 displays predictive margins (estimates of values [response means] predicted by the model) and standard errors of mutual-help group participation (attendance, involvement) and alcohol outcomes by attendance mode and time from models testing the interaction between timepoint and attendance mode in unadjusted and adjusted models. The interaction between time and mode in the adjusted model examining attendance was not significant. In addition, the interaction between time and mode was not significant in adjusted or unadjusted models examining involvement.

Figure 1. Unadjusted Predicted Probability (with 95% CI) of Attending 8+ Meetings in the Past 30 Days by Attendance Mode and over Time.

Notes. This figure depicts the predicted probabilities computed from the unadjusted model testing the interaction between mode and time (and its component terms) on attending 8+ meetings. The Wald Test for the joint effect of the interaction was statistically significant; χ²(4)=10.01, p=0.040.

Table 3.

Predictive Margins for Mutual-Help Group Participation and Alcohol Outcomes by Time and Attendance Mode.

	Unadjusted Models			Adjusted Models
	Baseline	6 months	12 months	Baseline	6 months	12 months
	Margin (SE)	Margin (SE)	Margin (SE)	Margin (SE)	Margin (SE)	Margin (SE)
Attendance Characteristics
8+ Meeting Attendance
In-person only	0.35 (0.05)	0.33 (0.06)	0.46 (0.07)	0.34 (0.05)	0.28 (0.05)	0.38 (0.06)
Online only	0.41 (0.03)	0.39 (0.03)	0.41 (0.03)	0.43 (0.03)	0.42 (0.03)	0.43 (0.03)
Both in-person and online	0.62 (0.03)	0.71 (0.04)	0.61 (0.04)	0.61 (0.03)	0.70 (0.04)	0.64 (0.04)
Involvement
In-person only	0.67 (0.03)	0.75 (0.02)	0.77 (0.03)	0.67 (0.03)	0.73 (0.02)	0.75 (0.03)
Online only	0.55 (0.01)	0.59 (0.01)	0.57 (0.02)	0.57 (0.01)	0.60 (0.01)	0.58 (0.02)
Both in-person and online	0.68 (0.02)	0.71 (0.02)	0.69 (0.02)	0.67 (0.02)	0.71 (0.01)	0.69 (0.02)
Alcohol Outcomes
Alcohol Abstinence
In-person only	0.59 (0.06)	0.68 (0.05)	0.64 (0.06)	0.56 (0.06)	0.65 (0.06)	0.61 (0.05)
Online only	0.48 (0.03)	0.59 (0.03)	0.66 (0.03)	0.51 (0.02)	0.61 (0.03)	0.67 (0.02)
Both in-person and online	0.49 (0.03)	0.55 (0.03)	0.57 (0.03)	0.50 (0.03)	0.56 (0.03)	0.61 (0.03)
None	---	0.55 (0.04)	0.52 (0.04)	---	0.56 (0.04)	0.52 (0.04)
Heavy Drinking Days
In-person only	2.34 (0.82)	0.32 (0.21)	0.98 (0.54)	5.95 (2.60)	0.63 (0.46)	1.43 (0.81)
Online only	2.61 (0.35)	0.91 (0.21)	1.26 (0.29)	3.85 (0.78)	1.29 (0.37)	1.66 (0.47)
Both in-person and online	1.65 (0.32)	1.12 (0.29)	1.25 (0.27)	2.67 (0.73)	1.74 (0.57)	1.24 (0.32)
None	---	4.00 (0.97)	2.52 (0.71)	---	4.86 (1.33)	2.26 (0.70)
Alcohol Problems
In-person only	0.50 (0.06)	0.21 (0.05)	0.28 (0.05)	0.52 (0.05)	0.25 (0.05)	0.31 (0.05)
Online only	0.57 (0.03)	0.35 (0.03)	0.26 (0.03)	0.57 (0.03)	0.34 (0.02)	0.27 (0.03)
Both in-person and online	0.57 (0.03)	0.38 (0.03)	0.38 (0.03)	0.57 (0.03)	0.38 (0.03)	0.35 (0.03)
None	---	0.34 (0.05)	0.31 (0.05)	---	0.34 (0.04)	0.31 (0.04)

Notes. All participants attended mutual help groups at baseline per study eligibility criteria, but models examining the time by mode interaction on attendance characteristics were run only on those who reported attending mutual help groups at follow-up. Models looking at alcohol outcomes were run irrespective of reported mutual help group attendance at follow-up.

3.3. Associations between attendance mode and alcohol outcomes

Table 4 provides results from unadjusted and adjusted models examining associations between mode and time with alcohol outcomes, which included the groups of participants who did not attend meetings in the 30 days prior to follow-ups. In the unadjusted models, there were significant mode associations indicating lower odds of abstinence (OR=0.54, p=0.006) and more heavy drinking days (IRR=2.50, p=0.005) for participants no longer attending mutual-help groups compared to those attending in-person only. In the unadjusted models, participants attending both in person and online had more alcohol problems than participants attending in person only (OR=1.59, p=0.022). In the adjusted models, while mode was not significant for alcohol abstinence or alcohol problems, it remained significant for heavy drinking days such that no longer attending meetings was associated with more heavy drinking days compared to attending in-person meetings only (IRR=2.23, p=0.048).

As Table 4 also displays, in the unadjusted models, there were significant (p<0.05) associations for time on all alcohol outcomes, suggesting improvement between baseline and each follow-up on increased odds of alcohol abstinence, decreased heavy drinking days, and decreased odds of alcohol problems. The associations remained significant in the adjusted models.

One significant interaction was found between time and mode with respect to alcohol outcomes, which was for heavy drinking days in the unadjusted model (Wald Test 2(5)=13.07, p=0.023; model coefficients not shown). Predictive margins in Table 3 and graphed in Figure 2 illustrate that days of heavy drinking over time stayed about the same among participants attending both online and in-person meetings; decreased and then increased between 6 and 12 months among participants attending online-only or in-person-only meetings; and declined from 6 to 12 months among participants with no attendance. In the adjusted model, the interaction of time and mode for heavy drinking days was not significant. In addition, the interaction of time and mode for abstinence and alcohol problems was not significant in unadjusted or adjusted models.

Figure 2. Unadjusted Linear Prediction (with 95% CI) of Number of Heavy Drinking Days in the Past 30 Days by Attendance Mode and over Time.

Notes. This figure depicts the exponentiated linear predictions computed from the unadjusted model testing the interaction between mode and time (and their component terms) on heavy drinking days in the past 30 days. The Wald Test for the joint effect of the interaction was statistically significant; χ²(5)=13.07, p=0.023.

4. Discussion

This study of mutual-help group participation among people with lifetime alcohol use disorders found that online meeting attendance, alone or in combination with in-person attendance, was associated with a higher likelihood of frequent attendance compared to in-person attendance only. This is consistent with previous studies (Galanter et al., 2022; Hoffman & Dudkiewicz, 2021) and with the observed greater accessibility and convenience of online meetings (Senreich et al., 2022). Online mutual-help group meetings may have potential for reaching and supporting people who might not otherwise engage in treatment or other sources of support (Beck et al., 2023a,b).

However, the present study also found that attending online meetings exclusively was associated with less mutual-help group involvement compared to attending meetings exclusively in-person. This finding aligns with previous studies suggesting that online meetings gave attendees a lower sense of shared alliance than in-person meetings (Barrett & Murphy, 2021; Hoffman & Dudkiewicz, 2021; Senreich et al., 2022). Meeting online rather than in-person may make it more difficult to build program practices tapped by involvement as measured here, such as socializing with other members and developing the closeness entailed by sponsorship or friendship. That less involvement was associated with online attendance only is important because involvement appears to be as or more important than meeting attendance for better alcohol outcomes, in that involvement remains a significant predictor of better outcomes even after accounting for attendance (Kaskutas et al., 2005; Timko et al., 2013; Timko et al., 2023). Further, in a systematic review of 12-step groups, involvement, as well as attendance, was negatively correlated with symptom severity and positively corelated with quality of life (Leurent et al., 2023). Beck et al. (2023b) noted that in order to maximize the benefits of online meetings, further research is needed on how best to support people to develop a change plan within the online group setting. Such support could emphasize aspects of group involvement related to growing relationships that foster recovery. In the present study, unlike participants with online attendance only, those who participated in both online and in-person meetings had comparable involvement to in person-only attendees.

That participants attending meetings online only was (at baseline) the largest group, whereas those attending meetings in-person only was the smallest, suggests that virtual attendance may be a lasting effect of the pandemic. Even though this study’s baseline recruitment was conducted almost a year after COVID-19 vaccines became available and the number of US cases and deaths had started to fall, online attendance predominated. The large online-only group further highlights the necessity of additional research to ensure not only attendance but also involvement among people attending meetings exclusively online. Meanwhile, people attending online meetings, alone or together with in-person attendance, were less likely than in person-only attendees to have the alcohol-related goal of total lifetime abstinence. Indeed, about one-half of online-only and of dual online and in-person attendees did not want to abstain forever at baseline, compared to about one-third of in-person-only attendees. Such findings add urgency to the observed need for psychosocial treatment of substance use disorders that is not exclusively abstinence-based as a potential way to facilitate early initiation of treatment, engage and retain more people in treatment, and improve treatment effectiveness (Paquette et al., 2022).

The present study also found that although online-only attendance was associated with less mutual-help group involvement, online-only attendees had comparable alcohol outcomes to in-person-only attendees across surveys. Perhaps this was due to the small size of the in-person only group, which may have limited statistical power. In addition, online-only attendees had (on average) relatively more protective factors at baseline, including having a spouse or partner, a post-college education, an income in the $80-$120,000 range, and a lower likelihood of drug problems. Dual online and in-person attendance was also associated with comparable alcohol outcomes to in person-only attendance. In keeping with studies supporting the benefits of sustained mutual-help group participation (Adelman-Mullally et al., 2021; Tonigan et al., 2018), the present study found that ceasing to attend mutual-help groups in the 30 days prior to follow-ups was associated with a higher likelihood of heavy drinking, but not with less abstinence or more alcohol problems, compared to attending mutual-help groups in-person.

4.1. Limitations

One limitation is that the present study could not establish a baseline response rate and ensure the representativeness of the mutual-help group sub-samples (Timko et al., 2022; Zemore et al., 2017). It was not possible to establish a baseline denominator or to compare baseline responders to non-responders, such that sample selection biases cannot be ruled out. In particular, the sample may be biased in that all surveys were completed online. Participants may have been more knowledgeable about or comfortable with using online resources than mutual-help group attendees generally, which may limit the generalizability of results. Another concern is that participants were combined across 12-step programs, although members of these programs differ on demographic, clinical, and other characteristics, and the programs vary on size and meeting availability. However, combining 12-step participants was necessary due to the rather small samples of groups other than AA. Limitations also include that participants did not provide detailed information on alcohol treatment they may have received. Participants reported at baseline whether they had received inpatient or outpatient treatment for an alcohol or drug problem in the past year, and at follow-ups whether they had received such treatment in the past 6 months. Although including an indicator of treatment receipt to the adjusted models predicting mutual-help group participation and outcomes did not change the findings, more fine-tuned analyses of treatment experiences are warranted. Together, these limitations underscore the importance of not interpreting statistically nonsignificant comparisons or associations as sufficient evidence to conclude there is no difference between groups. Nonsignificant p values should not be regarded as support for the null hypothesis; rather, p values larger than the threshold indicate only that the statistical test was incapable of rejecting the null hypothesis (Aczel B et al., 2018). This could have occurred because the association tested does not exist, but also because the power of the test was insufficient to detect a true effect (Aczel et al., 2018). The present study’s limitations also underscore the need for additional studies to further inform the present findings, such as combined analyses of the PAL Study 2015 and 2021 Cohorts.

4.2. Conclusions

Despite its limitations, this study advances the evidence base on alcohol use disorder recovery by suggesting that online mutual-help group attendance is associated with attending more meetings but also with less involvement and lower endorsement of abstinence as a recovery goal. Among survey respondents, all of whom had attended a mutual-help group meeting prior to baseline, meeting attendance did not increase at follow-ups, but involvement grew and alcohol use outcomes improved during the follow-up period. These results augment findings that the majority of individuals who develop alcohol use disorders reduce or resolve their problem over time (Tucker et al., 2020). During recovery, goals often change from a non-abstinence toward a complete-abstinence goal, and people who seek social support, such as that available through mutual-help groups, are more likely to change their goal (Eddie et al., 2022; Schwebel et al., 2022). Although abstinence is not necessary to overcome a substance use disorder, it is likely to lead to better functioning and greater well-being (Eddie et al., 2022). To help researchers and clinicians working with people seeking recovery with different goals, next steps are to determine if the present study’s findings are replicated, and if so, to better elucidate the consequences of less involvement among online-only attendees, and to determine mechanisms that explain alcohol outcomes among mutual-help group attendees utilizing different modes.

Highlights.

• In-person-only attendees more often had the goal of lifetime alcohol abstinence.

• Online-only attendance was associated with more meetings but less involvement.

• Involvement and alcohol outcomes, but not attendance, improved at follow-ups.

• Attendance mode did not relate to alcohol outcomes.