OUP user menu


J. H. Foster, E. J. Marshall, T. J. Peters
DOI: http://dx.doi.org/10.1093/alcalc/35.6.580 580-586 First published online: 1 November 2000


Research has tended to show that the gains of residential rehabilitation are short-term and cost-inefficient. This study compares the outcomes of two samples, one group staying at a non-statutory sector alcohol detoxification unit for ≤7 days (short stay: SS) with a second group also admitted for detoxification but who stayed at the Unit for a further 8–21 days (long stay: LS). Allocation was not at random: the longer stay was either at the request of the client, referring or treatment agency itself and then had to be approved by an external funding agency. Sixty-four DSM-IV (Diagnostic and Statistical Manual of Mental Disorders) alcohol-dependent subjects were studied. Baseline data included socio-demographic information, illicit drug use during the past 12 months, severity of alcohol dependence, alcohol problems, physical/psychological symptoms, depression and indices of quality of life. At baseline, LS subjects reported more recreational cannabis use than SS subjects. Sixty-two (97%) subjects were re-interviewed 12 weeks after baseline assessment. During follow-up, equal proportions of each group relapsed (≥21 units/7 day period for males; ≥14 units/7day period for females). There was a trend for SS clients to have consumed less alcohol in total than the LS clients. The trend was towards improvement in the study measurements for the SS group, though none of the changes was significant. In the LS group, all variables tended towards a deterioration in health status. The longer stay did not appear to confer any extra benefit to the LS group. Cannabis use and illicit drug use at baseline, while commoner in the LS group, did not predict drinking or social adjustment in the follow-up period in this sample and thus could not be used to explain the lack of a better outcome in the LS group.


The value of residential rehabilitation as a treatment for alcohol problems has been the subject of debate. Some authors have reported that any gains tend to be short-lived and cost-inefficient (Holder et al., 1991; Hodgson, 1994; Finney and Monaghan, 1996). However, after reviewing the literature on duration and intensity of residential care in an editorial for the Journal of Studies on Alcohol, Schuckit (1998) felt that the existing data did not justify a move away from in-patient treatments, providing that treatment ‘intensity’ (Finney et al., 1996) is maintained. Circumstances where in-patient treatment is indicated include medical or psychiatric disorders impacting on an individual's capacity to benefit from alcoholism treatments, an environment making out-patient treatment unsafe, homelessness, living with other addicted individuals, chaotic or dysfunctional family environment, or a history of difficulty in abstaining in an out-patient setting (Nace, 1993).

There have also been a number of controlled studies that found no significant differences in outcome between treatments of differing durations. These were reviewed by Miller and Hester (1986) who concluded that longer or more intensive programmes failed to produce any significant benefits when compared to those that were shorter and less intensive.

In the UK, the National Health Service and Community Care Act (1990) transferred the budget for funding non-National Health Service alcohol services to Local Authority Social Services. Our study took place after the Act was operational and evaluated 3-month outcomes for alcohol-dependent subjects who received residential treatment at a non-statutory sector alcohol detoxification Unit. This unit employed two treatment modes, a short stay (SS) and a longer stay (LS).

We compared outcomes of the SS with the LS programme. Two questions were addressed: first, whether the allocation of local authority funding correlated with certain social or clinical criteria determined at the time of the patient's admission; second, whether the longer stay conferred any benefits in terms of improved quality of life or other benefits after a 3-month period.


Subjects and general protocol

The SS group were admitted for detoxification only, which could last up to 7 days, i.e. predominantly a medical intervention. Alcohol withdrawal was treated with diazepam prescribed according to the following protocol: 10 mg 6 hourly for 24 h, 5 mg 6 hourly for 24 h, 2 mg 6 hourly for 24 h and then stop. This was initiated by a qualified first level nurse (general or psychiatric) when the pulse was ≥100 beats/min and the blood alcohol level (BAL) ≤60 mg/dl. There was a 24 h qualified nursing cover. Medical input was not routinely available on site, but part-time medical support was provided by local general practitioners (GPs). Residents attended the GP surgery as soon as possible after admission. Medical assessment consisted of a brief physical/psychological assessment lasting about 5 min. Unless otherwise indicated, or requested by the subject, this could be the only contact with a doctor during the SS. There were no specific on-call arrangements made between the GPs and the Unit. Throughout the SS, an allocated worker saw residents for at least 15 min daily. This meeting consisted of an enquiry into the progress of the detoxification, discussion of plans for discharge, whether the individual thought a longer residential period was necessary and any other issues brought up by the resident. There was no group programme during the SS period.

With funding approval from the individual's local authority, a further residential period after detoxification of up to 21 days was possible (LS). The longer stay was called an assessment period by the unit and aimed to allow the subjects to ‘assess their own needs’. Usually this meant using the additional time to set up further community or residential support. During the longer stay, the subject was allocated a key-worker who could be a nurse or an unqualified social worker, who assisted subjects to ‘assess their own needs.’ There were usually two key-worker sessions weekly. Attendance at a daily (Monday– Friday) group programme was expected. The group programme focused on relapse prevention and training in assertiveness and social skills. There were also house groups where issues concerning the general running of the Unit, such as a cleaning rota, were discussed. During the LS period, the requirement that an individual was seen by staff for 15 min per day ceased.

Subjects were eligible if they were DSM-IV alcohol-dependent subjects (American Psychiatric Association, 1994), able to provide informed consent, an address where they could be followed-up and were not DSM-IV opiate-dependent. The principal author (J.F.) established eligibility for the study.

Consecutive admissions to the unit between October 1995 and June 1997, who met eligibility criteria, were asked for their informed consent to participate in the research. Those who did not go on to the ‘assessment programme’ became the SS group (except three subjects who refused to grant informed consent and two who were sentenced to imprisonment whilst in residence and thus did not have the opportunity to drink during the follow-up period). Those who proceeded to ‘assessment’ became the LS group (except two subjects who refused consent). Thus 64 subjects, 32 in each group, were recruited. They were interviewed at intake and at 12 weeks after baseline assessment.

Forty-three (67%) of the subjects were self-referrals to the Unit, eight (12%) came from either a GP or local statutory community alcohol services respectively and five (9%) were either from the probation service or social services.

Measures at baseline only

Baseline data were collected 4–5 days after admission when subjects had completed their detoxification regime and were drug-free. This included age, ethnicity, marital status, living arrangements and social class. Social class was assigned in accordance with definitions contained in the Registrar General's classification of socio-economic status (Office of Population Censuses and Surveys, 1991). Present and last employment was the assigning variable for men, and for women employment of the spouse or father was applied if they were not working or had not worked themselves.

The Severity of Alcohol Dependence Questionnaire

The SADQ is a 20-item self-report questionnaire devised by Stockwell et al. (1979) to report the physical and affective symptoms of alcohol withdrawal, craving and relief drinking, typical daily alcohol consumption and re-instatement of symptoms after heavy drinking within the past 6 months. Each item can attain a score of 0, 1, 2 or 3; thus the maximum score is 60. A score of >30 is considered to indicate severe dependency (Stockwell et al., 1983).

Alcohol Problems Questionnaire (APQ)

This is a self-report measure comparing 44 items, which was devised to explore the relationship between alcohol problems and alcohol dependence over the previous 6 month period (Drummond, 1990). The domains relating to employment, marriage, and children were not completed as they did not apply to the majority of the sample. Thus, each subject completed 23 questions concerning alcohol problems in the following areas: friends, finances, police-related, physical and psychological difficulties. Every positive response was assigned a score of 1 and the range of possible scores was 1–23.

Baseline and follow-up measures

General Health Questionnaire (GHQ-12).

This is a 12-item self-administered questionnaire that measures non-psychotic psychiatric status over the past month (Goldberg, 1972). There are four possible responses to each question, which were scored 0–0–1–1. A score of ≥2.5 is the cut-off point for ‘psychiatric caseness’ (Banks, 1983).

Beck Depression Inventory (BDI).

This measures 21 items of depressive symptomatology over the previous 7 days (Beck et al., 1961). Scores >9 indicate clinical depression (Bowling, 1991).

Rotterdam Symptoms Checklist (RSCL).

This was designed as a measure of the physical and psychological symptoms of cancer present during the past 7 days (deHaes et al., 1990). It was chosen for this study, because many of the symptoms listed were also present in alcohol misusers during detoxification. These included lack of appetite, worrying, difficulty in sleeping, diarrhoea, shivering, tingling in the hands and feet and dry mouth. The range of scores is 30–120. Higher scores are indicative of a greater severity of symptomatology

Life Situation Survey (LSS).

This is a subjective measure of current quality of life (QoL), relevant to those with chronic illnesses and includes items on work, leisure, nutrition, sleep, social nurturance, earnings, health, love/affection, environment, self-esteem, security, public support, stress, mobility, autonomy, energy level, social support, mood/affect, outlook and egalitarianism (Chubon, 1987). The range of scores is 20–140, and scores of ≤90 denote a poor QoL.

Nottingham Health Profile (NHP).

This is a self-administered questionnaire, designed to measure ‘perceived health problems and the extent to which such problems affect daily activities’ (Hunt and McEwen, 1980). It contains items on: physical mobility, pain, sleep, social isolation, emotional reactions, and energy level. Higher scores reflect greater current disability.

Outcome-only indices

Indices measured at follow-up were relapse rates, days taken to relapse and units of alcohol consumed. A relapse was defined as consumption of at least 21 units of alcohol (males) and 14 units (females) over a 7-day period (8 g of alcohol = 1 unit). This accords with the recommended UK level of ‘sensible’ drinking (Royal College of Psychiatrists, 1986; Faculty of Public Health Medicine–Royal College of Physicians, 1991).


For categorical variables, differences between the SS and LS groups were tested by two-tailed χ2 or Fisher's exact analyses. Between-group univariate analyses for data of at least ordinal status were by two-tailed Mann–Whitney U-test or independent sample t-tests, dependent upon whether the data were non-parametric or parametric respectively. Multivariate tests were also used to assess between-group differences. Where the dependent variable was categorical, a forward stepwise logistic regression of all the baseline variables was used. A linear regression analysis with the number of days of admission as the dependent variable was also performed.

Within-group changes were examined by means of a two-tailed Wilcoxon or paired t-test, dependent upon whether the spread of the data was non-parametric or parametric respectively. If there were any significant results from the regression analyses, then these variables were controlled for in the within-group analyses. The method used was a repeated measures ANOVA entering the independent variable as the between-group factor. The dependent variable was log-transformed in the case of parametric data distributions.

Many of these results produced an abnormal distribution of the data; where this is the case the semi-interquartile range (SIQ) is recommended as the indicator for the spread of the data (Altman and Bland, 1994). This was used for the baseline data and the SEM for the within-group changes. The region within which statistical values led to the rejection of the null hypothesis was P ≤ 0.05 throughout. All analyses were carried out by means of the SPSS for Windows 6.1 computer package.


There were 64 subjects (40 males, 24 females) with a median age of 41 years (range 24–64). As shown in Table 1, the majority were single, divorced or separated, living alone, from a lower social class, and unemployed; no socio-demographic differences distinguished the groups, but the LS group tended to be more often unemployed (P = 0.054).

View this table:
Table 1.

Comparison of the socio-demographic features of the short stay and the long stay samples

Short stay (n = 32)Long stay (n = 32)P
aDenotes independent sample t-test; all other analyses were two-tailed χ2-tests unless <5 cells present in which case a two-tailed Fisher's exact test has been used. Significance was taken as P ≤ 0.05.
bValues shown are numbers with percentages in parentheses.
Short stay refers to those staying for ≤7 days; long stay refers to those subjects staying for >7 days.
Social class I/II: professional and managerial occupations; social class III: skilled and non-skilled manual workers; social class IV/V: unskilled manual/non-manual workers and never-employed workers.
Mean (SIQ) (years)43.5 (5)41.4 (6.2)0.347a
Male23 (72)17 (53)
Female9 (28)15 (47)0.121
White English19 (59)21 (66)
Scottish4 (12)2 (6)
Irish7 (22)7 (22)
Other2 (7)2 (6)0.857
Marital statusb
Single18 (56)15 (47)
Married3 (5)6 (19)
Divorced/separated11 (34)8 (25)
Widowed3 (5)3 (9)0.147
Living arrangementsb
Alone16 (50)16 (50)
Family unit13 (41)11 (34)
Other (institution etc.)3 (9)5 (16)0.376
Unemployed23 (72)29 (91)
Employed9 (28)3 (9)0.054
Social classb
I/II9 (28)7 (22)
III10 (31)11 (34)
IV/V13 (41)14 (44)0.845

A comparison of the number of admission days produced the following results: SS: mean 6.6, SIQ 0.1; LS: mean 25.1, SIQ 1.5. In the SS sample, four subjects (12%) did not complete the programme and the equivalent figure for the LS group was eight subjects (25%).

Clinical measures at baseline

LS and SS groups did not differ at baseline in terms of smoking, SADQ, APQ, BDI, GHQ-12, LSS, RSCL scores, NHP sub-/total scores, present/past alcohol use or prescription drug use. They tended to be moderately to severely dependent, as measured by the SADQ, 34 clients (53%) having scores in the severe range, >30 (SS: mean 30.1, SIQ 10; LS: mean 32.1, SIQ 8).

Twenty-eight subjects (44%) had used illicit drugs recreationally (by their own definition) over the 12 months prior to admission. Cannabis was the drug of choice for 25 (89%) of these subjects. No other category of illicit drug was used by more than four (14%) subjects. Cannabis use within the past 12 months was associated with a significant likelihood of being in the LS group (LS: n = 18, 56%; SS: n = 7, 22%) (P = 0.005). Non-recreational drug use was significantly associated with membership of the SS group (SS: n = 23, 72%; LS: n = 13, 41%; P = 0.012).

Recreational cannabis use, which was a categorical variable, emerged as the only significant differentiating variable using a forward stepwise logistic regression entering all the baseline variables (R = 0.248, SE = 0.291, df = 1, P = 0.005). Recreational cannabis users were 2.3 (OR = 2.3, 1.17–4.48, 95% CI) times as likely to be in the LS group as other subjects. In addition, a forward stepwise linear regression was performed using days of admission as the continuous dependent variable; again the only significant (P ≤ 0.05) differentiating variable was recreational cannabis use (beta = 0.265, SE beta = 2.613, t = 2.09, significant t = 0.041).

To test whether recent criminal history was a factor in greater resource allocation (i.e. funding given for long stay) the policerelated problems APQ sub-score was tested. This comprised three variables: trouble with the police, drinking and driving, and imprisonment; the possible range of scores was 0–3. Forty-three (67%) of the total had a score of zero (i.e. no police-related alcohol problems) and there was not a significant between-group difference (SS: mean 0.35, SIQ 0; LS: mean 0.48, SIQ 0.5; P = 0.467). A subsequent logistic regression was performed substituting the APQ police-related problem score for the APQ total score. The computer once again selected recreational cannabis use as the only significant variable.


Sixty-two individuals (97%) were contacted at 12 weeks, with two lost to follow up (both from the SS group). Of the 30 SS patients followed up, 20 (66%) had relapsed. Of the 32 LS patients followed up, 18 (56%) had relapsed. The difference was not statistically significant (P = 0.301). When the two SS patients lost to follow-up were categorized as having relapsed, the difference remained statistically insignificant (P = 0.226). In the patients available for follow-up, no significant difference emerged for days taken to relapse into heavy drinking (SS: mean 10.5, SIQ 5.6; LS: mean 17.7, SIQ 14; P = 0.414). If the two subjects lost to follow-up were assumed to have relapsed on their proposed day of discharge, i.e. day 7, the between-group differences were again not statistically significant (SS: mean 10.3, SIQ 4.8; LS: mean 17.7, SIQ 14; P = 0.116).

There was not a significant between-group difference in the number of units consumed during the study (SS: mean 635, SIQ 585; LS: mean 713, SIQ 630; P = 0.739). Of the 24 subjects who were followed up and known not to have relapsed, 14 (58%) were abstinent for the entire period. The remaining 10 drank alcohol at some point and these were termed controlled drinkers (CD). Two were engaged in a formal controlled drinking programme. Any other drinking took the form of a self-limiting lapse. There were five CD subjects in each group. However, the CD SS group subjects drank significantly fewer units of alcohol than the CD LS subjects over the study period (SS: mean 14.0, SIQ 12.8; LS: mean 63.5, SIQ 25.7; P = 0.036). The contributions of both cannabis use and illicit drug use to relapse, days to relapse, alcohol consumption and differences relating to controlled drinking were assessed. There were no significant associations (P > 0.1 on all occasions).

The within-group changes are shown in Table 2. None was significant. However, in the SS group, there were trends towards improvements in scores for eight variables. None of the measures showed a trend toward improvement in the LS group. Associations with illicit drug and cannabis use were assessed. Illicit drug use did not contribute significantly to any of the results (P > 0.05). Recreational cannabis use at baseline was significantly associated with two within-group findings. First, it was related to the improvement in NHP sleep sub-scores in the SS group (F = 6.47, df = 1,18, P = 0.020). Second, cannabis use was linked to the deterioration in depression scores as measured by the BDI in the LS group (F = 4.83, df = 1,29, P = 0.036).

View this table:
300 Multiple Choices

This is a pdf-only article and there is no markup to show you.


    This study investigated the 3-month outcomes of two groups of alcohol-dependent subjects who had received residential treatment at a non-statutory sector alcohol unit. Two variables at baseline differentiated the two groups, recreational cannabis and illicit drug use. The longer stay did not appear to confer any extra benefit. In fact, despite an extra 3 weeks of time in the wider community, the SS group had on average consumed less alcohol than the LS group. Gains in quality of life and psychiatric status tended to be seen more in the SS rather than the LS subjects.

    The interaction between cannabis use and alcohol dependency is under-explored (McBride, 1995). In the present study baseline cannabis use was not related to relapse or alcohol consumption outcomes; however, the worsening of depression scores in the LS group was linked to cannabis use. Subjects who have a psychological co-morbidity may be likely to self-medicate using cannabis. Thus, cannabis use in alcohol-dependent subjects could be an indicator of a greater array of problems. In contrast, cannabis use was associated with an improvement in sleep scores for the SS subjects.

    There was a trend approaching statistical significance for subjects from the LS group to be unemployed. McLellan et al. (1994) showed that unemployment was associated with poor outcomes regardless of the substance used, though this study focused upon opiate and cocaine use. A recently reported study (Curran and Booth, 1999) found that unemployment was a predictor of a poor 12-month outcome in a study of 298 male US veterans. Thus it is possible that LS subjects had a worse prognosis at baseline.

    Implications for UK local authority funders

    The two significant differentiating variables between the SS and LS groups were recreational cannabis and illicit drug use and thus the decision-making process of requesting and offering funding for a longer stay remains unclear. Our findings suggest that local authority funders are not providing additional resources to those subjects with a recent offending history, though the sub-group was small. Previous studies have shown that variables related to the LSS; social environment (Humphreys et al., 1997), physical symptoms (Mendelson et al., 1986) and mood/affect (Greenfield et al., 1998) predict outcome. Thus we would have expected them to have a greater impact upon the allocation of resources than they appear to do. Empirical data have shown that severity of alcohol-dependence does not predict outcome (Edwards et al., 1988). Social workers may be aware of these findings and thus it is not surprising that it did not differentiate the two groups.

    The study did not address the issue of the aims of community care assessors, but amongst these are likely to be the improvement in their client's drinking status and quality of life. Our findings indicate that, on many occasions, these goals were not achieved at 3-month follow-up. Community care assessors are usually qualified social workers and there is a strong likelihood that there are deficiencies in their knowledge concerning recovery from alcoholism. Harrison (1992) found in a nationwide survey of students undergoing a 2-year CQSW social work training course, that there was a median of 8 h devoted to substance misuse throughout their training. Since this study, social work training courses have substantially changed and there is a need to investigate whether this deficiency has been addressed. As a result of the present study findings, we have designed a questionnaire which will be administered to a sample of community care assessors in London. There are two main areas of inquiry, the level of training and preparation community care assessors are given to assess the needs of alcohol-dependent subjects, and those variables which influence funding decisions.

    Implications for the non-statutory sector alcohol unit

    It is possible that there were significant improvements in the LS group at the end of the 28-day assessment period, but this was not assessed in our study. Unfortunately, any such postulated gains had been lost at 3 months follow-up. There is a strong likelihood that subjects from the LS group were perceived to have greater treatment needs; unfortunately they do not seem to have been served by the additional 3 weeks. During the study, most outcome variables improved in the SS group, albeit not to a statistically significant degree, whereas, in all the variables for the LS group, there was a deterioration in scores.

    A review of studies comparing outcomes in in-patient and out-patient treatments (Finney et al., 1996) found that few outcome differences between the groups treated in the two ‘settings’ were reported, but superior results occurred when subjects received more ‘intensive’ treatments. The SS groups had a compulsory 15 min contact with a Unit worker daily. This did not continue when the clients moved into the LS group. Had it continued, the 15 min might have conferred extra benefits for the LS group. We recommend that daily individual or group contact with a worker trained in motivational interviewing (Miller, 1983) and relapse prevention (Marlatt and Gordon, 1985) be standardized into all alcohol programmes.

    The findings of a recently published multi-centre nationwide study from the USA examining psychosocial treatments in cocaine dependence may provide some avenues for future alcohol research (Crits-Christoph et al., 1999). The best outcomes were provided by a combination of group drug counselling and manual-guided intensive individual drug counselling. Individual counselling had the expressed aim of stopping current drug use. Interestingly, this was the cheapest of all the therapies tested and delivered for the shortest period, the others being cognitive therapy and supportive expressive therapy, which required specially trained therapists to carry them out.


    Our findings are dependent upon self-report data. Sobell and Sobell (1988) concluded that self-report data of alcoholics who are in treatment and given a guarantee of confidentiality accords closely with official records, biochemical markers and family reports. Many studies attempt to corroborate data by the use of collateral reports. The majority of our subjects lived alone and had minimal contact with their families. Thus, little reliability could be placed upon such reports. A limitation of the study is that, for the follow-up assessment, the interviewer was not always blind as to whether the subject was in the LS or SS group. This was often clear from comments inadvertently made by the interviewee. Subconsciously there may have been some interviewer bias.

    With only 64 subjects, there is insufficient statistical power to allow anything other than tentative conclusions. A larger sample may have revealed that the longer stay was advantageous. But there was no sign of a trend towards this; if anything it was in the other direction. However, there also remains the possibility that the LS group contained subjects with poor prognosis, because they had greater social dysfunction and poorer social resources. It is likely that a further residential period was requested for the LS group by the Unit staff, because such subjects were perceived as needing greater residential and community support.

    Despite being non-randomized and naturalistic, we believe that there are four reasons to suggest our findings are deserving of a wider audience. First, there has been a call for more naturalistic studies from eminent sources and a number have recently been published. Finney et al. (1996) in their seminal review of the ‘mediators and moderators of setting effects’ advocated more naturalistic studies on the grounds that pre-selection and random assignment of patients do not ‘mirror treatment selection processes in non-research situations.’ Of three recently published naturalistic studies (Ouimette et al., 1997; Long et al., 1998; Trent, 1998), two have examined the effect of reducing length of residential stay. Long et al. (1998), studying subjects recruited from an independent hospital based in the UK, found that a 2-week in-/day-patient programme was more cost-efficient than an in-patient programme lasting 5 weeks. Trent (1998) followed-up a sample of 2823 alcohol-dependent subjects from the US navy's alcohol treatment programme for 12 months. One group remained for 4 weeks and another 6 weeks. The author concluded that the programme could be standardized to 4 weeks without an adverse impact upon outcome.

    Second, our results are consistent with previous findings, in particular Long et al. (1998) and Trent (1998). Although our sample size was comparatively small, there were trends. The quality of life and mood measures for subjects from the SS group tended to improve, whilst those from the LS sample tended to deteriorate. Third, although our sample was not randomized, we have studied consecutive admissions. Fourth, at 97%, there was an impressive follow-up rate.

    What was responsible for the trends towards improvement (albeit not statistically significant) that occurred in the SS sample? There are two possibilities, the detoxification (plus daily contact with the worker) or the baseline assessment. If it was predominantly the detoxification and daily contact then our findings suggest that ‘treatment intensity’ should be maintained to promote good outcomes. This is especially so for subjects perceived to have greater treatment needs. The baseline assessment comprised an evaluation of a number of health states. As a result, cognitive dissonance (Miller, 1983) may have been created and this could be the input responsible for most of the changes. Ideally, the methodology should be designed to tease out the impact of baseline assessment.

    Performed in inner city London, there is a possibility that the findings concerning the policies of the local authorities were unique to this geographical area. However, we feel this is unlikely, as local authorities throughout the UK work to standardized procedures. Are findings from a non-statutory sector population generalizable to other statutory services? This is the first paper to evaluate a UK non-statutory sector residential service. Many of the subjects fit the profile of ‘revolving-door patients’ and will have also used statutory National Health Service facilities. In addition, local authority funding arrangements are not unique to the non-statutory sector. Subjects leaving National Health Service alcohol specialist treatment units wishing to have further residential rehabilitation are also subjected to the same procedures. For these reasons, we believe our findings are generalizable to other settings.


    John Foster was supported by a South Thames Regional Research and Development Training Fellowship. The statistical advice provided by Dr Richard Hooper is gratefully acknowledged. Special thanks are also given to all residents and staff at the Drink Crisis Centre, Crisis and Assessment Unit, Kennington, London, without whose co-operation this study would not have been possible.


    • * Author to whom correspondence should be addressed.


    View Abstract