OUP user menu


Michael Gossop , Jennifer Harris , David Best , Lan-Ho Man , Victoria Manning , Jane Marshall , John Strang
DOI: http://dx.doi.org/10.1093/alcalc/agg104 421-426 First published online: 12 August 2003


Aims: This study investigates the relationship between attendance at Alcoholics Anonymous (AA) meetings prior to, during, and after leaving treatment, and changes in clinical outcome following inpatient alcohol treatment. Methods: A longitudinal design was used in which participants were interviewed at admission (within 5 days of entry), and 6 months following departure. The sample comprised 150 patients in an inpatient alcohol treatment programme who met ICD-10 criteria for alcohol dependence. The full sample was interviewed at admission to treatment. Six months after departure from treatment, 120 (80%) were re-interviewed. Results: Significant improvements in drinking behaviours (frequency, quantity and reported problems), psychological problems and quality of life were reported. Frequent AA attenders had superior drinking outcomes to non-AA attenders and infrequent attenders. Those who attended AA on a weekly or more frequent basis after treatment reported greater reductions in alcohol consumption and more abstinent days. This relationship was sustained after controlling for potential confounding variables. Frequent AA attendance related only to improved drinking outcomes. Despite the improved outcomes, many of the sample had alcohol and psychiatric problems at follow-up. Conclusions: The importance of aftercare has long been acknowledged. Despite this, adequate aftercare services are often lacking. The findings support the role of Alcoholics Anonymous as a useful aftercare resource.

(Received 14 March 2003; first review notified 27 March 2003; in revised form 25 April 2003; accepted 5 May 2003)


Alcoholics Anonymous (AA) has a philosophy of mutual-help, group affiliation and identification (Spickard, 1990). AA plays a unique role in assisting the recovery of people with alcohol dependence disorders, and is widely believed to be an effective intervention for alcoholism (Miller and McCrady, 1993; Timko et al., 1994). It is the most widely sought form of help for those with alcohol problems in the USA (McCrady and Miller, 1993), and many treatments, especially in the USA, operate within the broad context of the Twelve-Step model (Wallace, 1996). AA may have about two million members world-wide (Makela, 1993). However, despite its popularity and impact on alcoholism treatment, relatively little controlled research into the benefits of Twelve-Step approaches has been conducted (Morgenstern et al., 1997; Kownacki and Shadish, 1999).

Although AA has claimed a 75% sobriety rate for drinkers who seriously invest in the programme (Alcoholics Anonymous, 1976; Thurstin et al., 1987), other sources estimated that 50% of AA participants drop out within the first 3 months (Chappel, 1993). However, increased abstinence has been reported among those attending AA following substance misuse treatment (Emrick, 1987; Christo and Franey, 1995; Humphreys et al., 1998; Ouimette et al., 1998; Fiorentine, 1999), and lower rates of relapse have been found to be associated with more frequent attendance at AA (Humphreys et al., 1997; Caldwell and Cutter, 1998; Fiorentine and Hillhouse, 2000). Affiliation with AA has been found to be more predictive of positive outcomes than attendance alone (McLatchie and Lomp, 1988; Miller and Verinis, 1995; Montgomery et al., 1995; Morgenstern et al., 1997).

In Project MATCH, those randomized to Twelve-Step Facilitation attended more AA meetings, and AA involvement was associated with better 3-year outcomes, specially for drinkers with social networks which were supportive of drinking (Longabaugh et al., 1998). Ouimette et al. (1997) also found that Twelve-Step and cognitive behavioural approaches were equally effective in maintaining abstinence from alcohol, and in reducing depression and anxiety.

Not all studies have found improved outcomes associated with AA attendance (McLatchie and Lomp, 1988; Miller et al., 1992; McCrady et al., 1996). The positive outcomes associated with AA attendance often declined after 6 or 12 months (Alford, 1980; Wells et al., 1994). Also, most AA research studies have been conducted in the US. Because of cultural differences in healthcare provision and treatment goals, this may not generalize to a UK context. While there are currently about 3500 AA meetings in the UK each week (Personal Communication with the AA General Service Board, 2003), little is known about AA’s role within UK substance misuse treatment services.

The present study is a longitudinal investigation of patients receiving treatment for alcohol problems within a specialist National Health Service (NHS) inpatient alcoholism treatment service. It reports patient outcomes for alcohol consumption, alcohol-related problems, psychological problems and subjective quality of life measures during the 6 months following treatment completion. In particular, it investigates the relationship between attendance at AA meetings prior to, during and after leaving treatment, and changes in clinical outcome.



The setting was a 16-bed, specialist, short-stay, inpatient unit for the treatment and rehabilitation of patients with alcoholism problems. Referral was predominantly from outpatient services across South London. The treatment programme combined a 10-day medically assisted, pharmacological detoxification (using chlordiazepoxide or oxazepam) with a therapeutic programme of individual and group work. Patients who wished to receive rehabilitation services were supported in this, and additional supports were provided regarding housing, state benefits and further treatment needs. AA meetings were held within the hospital premises but not specifically in the alcohol unit. Patients were encouraged to attend at least one AA meeting during their stay, but attendance was not compulsory and did not represent a core component of the therapeutic programme. In addition to meetings, members of a local AA group made regular visits to the inpatient unit.

Procedure and measures

There were no selection criteria other than alcohol dependence: all consecutive admissions to the inpatient unit were approached by an independent researcher (i.e. not part of the treatment team) and invited to participate. No data are available on refusal rates though this was low (less than 10%). A longitudinal design was used in which consenting participants were interviewed at admission (within 5 days of entry), and 6 months following discharge.

At each interview, participants completed a 60-minute, semi-structured interview, assessing drinking history, drinking behaviour, health and wider lifestyle issues. Assessment was made of frequency and quantity of alcohol use during the 30 days prior to each interview. Frequency of drinking was also assessed for the period since departure from the inpatient programme. A variable was created for ‘percentage drinking days’; this was calculated as the number of self-reported days drinking for the period between leaving the index treatment and the follow-up interview divided by the total number of days in the community (i.e. in a non-protected environment). The ‘Alcohol Problems Questionnaire’ (Drummond, 1990) was used to assess alcohol problems during the 6 month periods prior to admission to treatment and prior to follow-up. Alcohol dependence was assessed using the ‘Severity of Alcohol Dependence Questionnaire’ (Stockwell et al., 1979). The ‘Symptoms Checklist-53’ (Derogatis, 1993) was used to obtain measures of psychiatric problems (Global Severity Index, Symptom Total, Symptom Distress scores), and of caseness for nine primary symptom dimensions. Chubon’s (1995) ‘Life Situation Survey’ was used as a measure of quality of life.


Patients (n = 150) were interviewed an average of 2.1 (SD = 1.7) days following entry to the inpatient unit. All met ICD-10 criteria (World Health Organization, 1992) for alcohol dependence (F10.2). The sample subjects were contacted for follow-up by home visit, telephone and/or letter inviting them to complete a follow-up interview. Of the recruited sample, 120 (80%) were successfully contacted for the follow-up interview. Follow-up interviews took place, on average 159 days (5.3 months) following the patients’ discharge from treatment. Follow-up data were missing for 30 (20%) participants. This was for a number of reasons; two patients had died, five refused to participate further in the study, and 23 were uncontactable.


To test for possible differences between those who were followed up (n = 120; 80%) and those who were not followed up (n = 30; 20%), comparisons were made between these two groups. Both groups reported similar levels of pre-treatment drinking (frequency and amount), severity of alcohol dependence, number of alcohol problems, psychological well-being and quality of life. They also reported similar drinking and treatment histories. The only significant difference reported by the two groups was related to lifetime AA attendance. Those lost to follow-up reported attending significantly fewer AA meetings prior to treatment (mean ± SD, 12.2 ± 54.5) than those contacted for follow-up (mean ± SD, 38.4 ± 84.2; t = 2.08; d.f. = 68.3; P < 0.05).

To further examine the comparability of the two groups, the following variables were entered into a logistic regression to predict follow-up status: age at entry to treatment, pre-entry drinking (frequency and amount), severity of dependence, number of alcohol problems, psychiatric problems, quality of life, drinking and treatment history and previous AA attendance. While the model was significant at the 5% level (χ2 = 29.7; d.f. = 14; P < 0.05), none of the individual variables was statistically significant. The results suggest that those who were followed up were broadly similar to those who were not followed up in terms of pre-admission characteristics and problems.

Demographic profile and drinking problems at intake.

The sample consisted of 38 (25.3%) women and 112 (74.7%) men. Thirty-six (25%) were married or cohabiting, 57 (39%) were single and 53 (36%) divorced, separated or widowed. The majority were receiving social security/unemployment benefits (116; 80%) in the month prior to admission; 23 (16%) had been working, and six (4%) were retired (n = 145, valid demographic data). The majority (79; 54%) were living in rented accommodation, 35 (24%) in owned accommodation, 27 (19%) with friends or family, and one was receiving treatment in hospital.

The sample had been drinking for an average ± SD of 22.4 ± 10.2 years (range 2–51 years) and drinking problematically for an average of 11.5 ± 8.5 years (range 1–40 years). They reported first seeking alcohol treatment at a mean age of 37.4 years, an average of 4.8 years before entry into the current inpatient treatment and an average of 6.8 ± 7.2 years (range 0–40 years) after initiating problematic drinking.

At intake, the sample reported high levels of alcohol dependence (61% had SADQ scores of more than 30, denoting severe alcohol dependence), a large number of alcohol-related problems, high levels of psychiatric problems and poor quality of life.

Changes in alcohol consumption (for month prior to interview)

The proportion of participants who had consumed alcohol in the previous month fell from intake to 6-month follow-up: 148 (99%) were drinking in the month prior to treatment entry compared with 84 (70%) of those re-interviewed at 6 months. Half of the sample (61; 51%) reported drinking on a less frequent basis at follow-up, and more than two-thirds (82; 70%) reported a reduction in drinking amounts per day. These changes are also reflected in the reductions in mean frequency of drinking and in mean amounts of alcohol per day consumed at follow-up (see Table 1). There was also a reduction in the proportion who reported daily drinking, from 128 (86%) in the month prior to admission to 58 (48%) at follow-up (McNemar test, P < 0.001).

View this table:
Table 1.

Changes in alcohol consumption, alcohol problems, psychiatric problems, and quality of life (full sample on whom follow-up data available)

Among those who were drinking in the month prior to their follow-up interviews (84; 70%), there were significant reductions from baseline drinking in frequency (t = −4.0; d.f. = 83; P < 0.001) and quantity (t = 2.7; d.f. = 81; P < 0.01) of alcohol consumption. Those who were drinking during the month prior to the follow-up interview reported drinking an average of 27.1 ± 20.9 units on an average of 24.6 ± 9.1 days in the previous month.

Time to first lapse and ‘percentage drinking days’ (during 6-month follow-up period)

More than three-quarters of the sample (95; 79%) reported consuming alcohol at some point during the 6 months following their inpatient treatment. Average time to first lapse was 44.5 ± 57.7 days. For about half of those reporting a lapse to alcohol (48%), this occurred within 3 weeks of their departure from the inpatient treatment unit.

Due to individual variations in the time between leaving treatment and the follow-up interviews, the variable ‘percentage drinking days’ was calculated to represent the relationship between actual drinking and days available for drinking during the period between treatment departure and follow-up. During this time, 25 (21%) had not consumed alcohol: 31 (26%) had consumed alcohol on up to a quarter of available drinking days: 25 (21%) on between a quarter and three-quarters of available drinking days; and 38 (32%) on more than three-quarters of all available days, of whom 11 reported drinking on a daily basis during the entire follow-up period.

Attendance at AA

Nearly three-quarters of the intake sample had attended at least one AA meeting prior to their current inpatient episode, and just over half had attended in the previous year (see Table 2). Forty percent (n = 48) attended AA during the follow-up period, of whom seven (15%) had never attended AA prior to the index treatment episode. Of those attending AA during the follow-up period, 35% (17/48) attended on a weekly (or more frequent) basis, and 65% (31/48) on a less than weekly basis. Only one participant reported attending AA’s recommended 90 meetings in 90 days.

View this table:
Table 2.

Attendance at Alcoholics Anonymous meetings

A statistically significant positive association was found between the number of AA meetings attended during the inpatient treatment episode and attendance during the follow-up period (r = 0.68; P < 0.001). No relationship was observed between pre-treatment/lifetime attendance at AA meetings and either frequency of AA attendance during treatment (r = 0.04; P = 0.66) or post-treatment attendance (r = 0.01; P = 0.84). There was a small but not statistically significant correlation between AA attendance in the year prior to admission and attendance during treatment (r = 0.17; P = 0.06). No correlation was found between AA attendance in the year prior to admission and after treatment (r = 0.08; P = 0.36).

AA attendance and outcomes

Those attending AA during the follow-up period reported drinking on a less frequent basis than did non-AA attenders, both in the month prior to follow-up (AA attenders: 14.2 ± 14.0 days; non-attenders: 19.2 ± 13.1 days; t = 2.0; d.f. = 118; P < 0.05), and during the entire follow-up period (AA attenders: mean = 31 ± 34.8% drinking days; non-attenders: mean = 53 ± 40.7% drinking days; t = 3.9; d.f. = 110.6; P < 0.01). Among the follow-up sample, there were no significant differences between AA-attenders and non-attenders at follow-up in terms of drinking frequency (t = 1.9; d.f. = 118; P = 0.06) or changes in daily drinking amounts (t = 1.9; d.f. = 115; P = 0.06), although differences approached the 5% level of statistical significance for each of these variables.

In order to conduct more detailed analyses of these results with regard to frequency of AA attendance, three attendance groups were constructed: those who did not attend AA after treatment, infrequent attenders (less than weekly post-treatment attendance at AA), and frequent attenders (more than weekly). There were no significant differences between the three AA attendance groups in terms of the following pre-intake characteristics: amounts of alcohol consumed per day, severity of alcohol dependence, alcohol problems, psychiatric problems and number of other treatments received during the follow-up period (see Table 3). There was a difference with regard to drinking frequency at intake. While not significantly differing from those not attending AA during the follow-up period, those who attended AA on a weekly or more frequent basis reported less frequent drinking at treatment entry than those attending AA on a less than weekly basis.

View this table:
Table 3.

Intake characteristics of the three Alcoholics Anonymous follow-up attendance groups

After leaving the inpatient treatment service, and during the follow-up period, those who attended AA on a weekly or more frequent basis reported drinking less frequently and in lower amounts. Weekly or more frequent AA attenders also reported the greatest reductions in percentage drinking days. There was no relationship between frequency of AA attendance in the period following discharge and the number of days to first drink (lapse), nor changes in psychiatric problems, alcohol problems or quality of life (see Table 4). A small but statistically significant negative correlation was found between percentage drinking days and number of AA meetings attended during the follow-up period (r = −0.20; P < 0.05).

View this table:
Table 4.

Frequency of Alcoholics Anonymous (AA) attendance and outcomes

Further analyses of outcome were conducted to control for pre-intake characteristics, length of treatment stay and post-discharge treatment contacts. This allowed identification of variables that might confound the relationship between frequency of AA attendance and drinking outcomes, pre-intake factors and treatment differences between the three AA attendance groups.

Those variables that were found to differentiate between the three AA follow-up groups (baseline drinking frequency and days in treatment at the inpatient unit), and the number of additional treatment services contacted during the follow-up period (which was of borderline statistical significance) were entered as covariates in subsequent univariate analyses of covariance (ANCOVA). This did not alter the significant relationship between frequency of AA attendance and outcome drinking frequency. After controlling for baseline percentage drinking days, days in treatment and further treatment contacts, frequency of AA attendance was negatively associated with percentage drinking days during the follow-up period (F = 5.2; P < 0.01). A similar negative relationship was found between frequency of AA attendance and number of drinking days in the month prior to the follow-up interview (F = 4.3; P < 0.05).


Improvements, both in drinking and in other problems, were reported after inpatient alcoholism treatment. Although some improved outcomes were associated with AA attendance per se, post-treatment reductions in alcohol consumption were more strongly associated with frequency of post-treatment attendance at AA. Those patients who attended AA meetings on a weekly or more frequent basis after treatment reported greater reductions in drinking than did non-attenders and infrequent AA attenders. The relationship between frequency of AA attendance and reduced alcohol consumption was sustained after statistically controlling for potential confounding variables, such as baseline characteristics and length of stay in the inpatient unit.

The benefits associated with AA attendance were limited to a relatively small proportion of patients. Less than half of the sample reported post-treatment AA attendance, and only 15% reported attending AA meetings on a weekly or more frequent basis after treatment. However, the relationship between improved drinking outcomes and frequent AA attendance mirrors that found by Fiorentine (1999). The consistency of these findings is interesting in view of the differences in sample characteristics, follow-up times and measurement methods in these two studies. In comparison to Fiorentine (1999), the present study reported relatively low levels of AA attendance and affiliation.

Although improvements in alcohol consumption, alcohol-related problems and in lifestyle domains were reported by the sample as a whole, the association between enhanced outcomes and frequent AA attendance related only to drinking outcomes. Frequent AA attendance was not associated with reduced psychiatric problems nor with improved subjectively reported quality of life. It is possible that changes in drinking may occur prior to changes in other areas of functioning, or that a 6-month follow-up period covers too short a time period to observe a relationship between AA attendance and the lifestyle domains. It is also possible that psychological health and quality of life problems which were previously masked by heavy alcohol use become more salient during periods of sobriety. Whatever the interpretation of this finding, and despite the improvements in drinking observed at follow-up, the psychological health of many of the sample was still poor with average symptom scores remaining higher than norms for adult psychiatric inpatients (Derogatis, 1993).

As inpatient alcohol treatment services tend to see more complex and more severely problematic cases (Weiss, 1999), it is unrealistic to expect short-term treatment interventions to produce and maintain improvements in the absence of additional supportive care, and the support provided by AA may not be appropriate or sufficient to produce change in such problems. Despite the overall improvements, the continued psychiatric problems, generally poor quality of life and relatively high rates of problematic drinking at follow-up indicate the need for more comprehensive treatment input and aftercare.

This is the first study of an NHS programme to present such data. The study has a number of limitations. Patients were not randomly allocated to AA and non-AA conditions. The naturalistic design may have allowed some confounding of results due to a selection bias in the characteristics of those who attended AA, though post hoc control of intake variables suggested few differences between AA attenders and non-attenders. Also, as in other longitudinal studies, a certain number of participants were lost to follow-up, though the follow-up rate of 80% represents a satisfactory achievement. Although the patients who were recruited to the study but not interviewed at follow-up reported attending fewer AA meetings prior to treatment, when statistical controls were made for pre-treatment drinking behaviour and other problems, no relationship was found between pre-treatment AA attendance and outcomes at follow-up. Nor was there any relationship between pre-treatment and post-treatment attendance at AA meetings.

Future research might consider the use of longer follow-up periods, and more detailed investigation of motivation and behaviour during the period following treatment. Greater consideration of process issues is also required, with more detailed investigation of AA involvement, the temporal relationship between meeting attendance, lapses and other treatment attendance, and an exploration of the mechanisms through which AA attendance supports or enhances drinking outcomes (Moos et al., 1990; Montgomery et al., 1993; Timko et al., 1994, 1995; Tonigan et al., 1995, 1996; Finney and Moos, 1996).

The importance of aftercare has long been acknowledged. Despite this, most treatment systems continue to suffer from a marked lack of adequate aftercare services. The findings suggest that AA can provide a useful aftercare resource and that regular contact with AA may help to maintain the benefits initially accrued from alcohol treatment programmes. Regular AA attendance, particularly in a group where members feel supported, comfortable and among like-minded people, may provide a means of sustaining the gains obtained during inpatient alcohol treatment. Closer liaison between NHS treatment units and local AA groups could be encouraged, with sponsor visits to the treatment service and members from local AA groups contacting patients at the time of leaving treatment. However, some features of Twelve-Step treatment tend to be more acceptable than others to alcoholics (Best et al., 2001), and AA affiliation will not appeal to, or be a feasible option for all patients.


We thank the staff and patients who supported and participated in the study. The study was funded by the Eva and Hans Rausing Trust through the charity Action on Addiction.


  • * Author to whom correspondence should be addressed at: National Addiction Centre, Maudsley Hospital/Institute of Psychiatry, 4 Windsor Walk, London SE5 8RF, UK. E-mail: m.gossop{at}iop.kcl.ac.uk


View Abstract