Association of Therapeutic Communities Home Page

Briefing Paper

Therapeutic Community Effectiveness.
A systematic international review of therapeutic community treatment for people with personality disorders and mentally disordered offenders
Janine Lees, Nick Manning & Barbara Rawlings
School of Sociology and Social Policy, University of Nottingham, University Park, Nottingham NG7 2RD 1999
e-mail: nickmanning@nottingham.ac.uk

For the full published report, contact:
NHS Centre for Reviews and Dissemination
University of York
York YO1 5DD
United Kingdom

The structure of the review

This systematic literature review was commissioned by the High Security Psychiatric Services Commissioning Board (HSPSCB), to look at therapeutic communities in psychiatric and other settings, particularly for people with personality disorder. (This review was NOT intended to address issues of defining personality disorder per se.) A therapeutic community is "a consciously-designed social environment and programme within a residential or day unit in which the social and group process is harnessed with therapeutic intent. In the therapeutic community the community is the primary therapeutic instrument." (Roberts 1997, p. 4)

The original bid call cited Rapoport's (1960) four general principles for defining a democratic therapeutic community, and although this report addresses the various forms of the therapeutic community, the focus of this research review has been on existing literature relating to the democratic therapeutic community, in the tradition of Tom Main and Maxwell Jones, in both psychiatric and other secure settings, and in nonsecure psychiatric settings, and on those dealing specifically with people with personality disorders, and mentally disordered offenders.

The call for bids asked specifically for a review of the international literature on the effectiveness of therapeutic communities in such settings, and with such client groups. This review has therefore concentrated on the research literature on effectiveness, and the main part of the report concentrates on post-treatment outcome studies of secure, and non-secure democratic therapeutic communities, for people with personality disorders, and mentally disordered offenders.

Because of their presence in the research literature, we also included hierarchical, or concept- based therapeutic communities. These are usually for substance abusers, in both secure and non-secure settings, both psychiatric and non-psychiatric, although there are considerably more in the USA, Canada, and other parts of the world than in the UK.

In addition to the research literature, the researchers agreed to look at additional ways of finding research literature, and descriptive information about therapeutic communities internationally. This was done by targeting "grey" literature, and by identifying as many therapeutic communities nationally and internationally as possible These were subsequently written to, along with known writers, or workers in this field, asking for any published or unpublished research they might have, and, if possible, for information about their therapeutic community; and its principles, organisation and practices. Such a wide trawl also helped to reduce any possible publication bias in favour of positive results.

This work was conducted, in accordance with Centre for Reviews and Dissemination guidelines, in an explicit and structured manner, with clearly stated research objectives, and protocols guiding the work and criteria for describing the relevance and quality of identified research. The results of this literature review are presented in both narrative form and a meta-analysis. This is because the quality of data presented, and the analyses in the studies retrieved, enabled systematic meta-analysis only in part, while much of the literature was not numerically comparable.

The therapeutic communities

Most secure therapeutic communities admit male offenders only. Some concept-based therapeutic communities in women's prisons in the US are reported in the literature, usually as part of a large study which includes both men and women. Secure therapeutic communities are located mainly within prison and correctional services. Of these, only HMP Grendon in Buckinghamshire, England, is an entirely therapeutic community prison. Other therapeutic communities comprise small units inside larger mainstream prisons, although some German Social Therapeutic Institutions have been established in separate secure premises outside prisons.

Most democratic prison therapeutic communities specialise in personality disorders and recidivism, whilst concept-based therapeutic communities are directed specifically at substance abuse, which usually refers to drug rather than alcohol use. However, there are overlaps here, since recent studies of concept-based therapeutic communities in the community have suggested that there is a high level of co-morbidity between drug abuse and personality disorder, and between drug abuse and mental illnesses.

The client information for non-secure democratic TCs varies according to the service context, is diverse, and not easily summarised, so is given in detail in the review.

The service contexts described in the research literature on democratic non-secure settings are very varied. Many are located within the NHS, and often at the tertiary level of provision. Abroad, the hospital based therapeutic communities are described typically as part of a psychiatric hospital.

Non-secure therapeutic communities such as Henderson Hospital, might typically include psychopaths, sociopaths, personality disorders, and character disorders. For example, in the Henderson Hospital, the majority are young people, with a lower age limit of 18. 87 per cent of residents meet DSM-IV-R criteria for borderline personality disorder, and 95 per cent met criteria for at least one Cluster B Axis II diagnoses.

Treatment is usually voluntary. In secure therapeutic communities inmates are generally selected by staff. In other therapeutic communities selection is by the community, or by staff-patient assessment group. Inmates can leave if they choose to do so, or be expelled from the community for their behaviour.

All units offer a daily or community meeting, democracy or patient participation in decision- making and running the therapeutic community, and a predominance of group activities.

Concept therapeutic communities are organised very differently from democratic therapeutic communities. Thus is a huge body of literature on concept-based therapeutic communities and their outcomes. Although not the main thrust of this report, the relevant in-treatment and post- treatment outcome studies on the effectiveness of secure concept-based therapeutic communities are analysed, while the non-secure concept-based outcome literature is also summarised. This is provided both for information, and for contrast.

The findings

We began with 8,160 articles and other literature, reduced to 294 broadly covering the relevant area. 181 individual TCs were named in the literature found, in 38 different countries, of which most were in the UK or the USA.

For our core focus on personality disorders/mentally abnormal offenders, there were 52 items on outcome studies of secure democratic TCs, 41 on outcome studies of nonsecure TCs, and 20 items on outcome studies of secure concept-based TCs. There were only 10 RCTs of any sort, and 10 cross-institutional or comparative studies, and a further 32 studies using some kind of control. If we take the latter as the minimum level of rigour that is acceptable, then there were in total 52 acceptable studies, all of which are discussed in some detail at some point in the report. Of these 52, 41 relate to democratic type therapeutic communities.

A meta-analysis was set up for the 52 studies with controls. 23 studies were excluded where the outcome criteria were unclear, where the raw numbers were not reported, or where the original sample was net clearly specified before attrition. Where there was a choice of outcome measures and control groups, emphasis was placed on conservative criteria, such as reconviction rates rather than psychological improvements, and on nontreated controls. This reduced the number of studies for the meta-analysis to 29.

The analysis had two stages. Initially the odds-ratios for the individual studies, and 95% confidence intervals, were calculated (Woolf , 1955, discussed in Kahn and Sempos, 1989, pp. 56-57). Subsequently, the odds-ratios were combined to produce a summary odds-ratio for the 29 studies, and subsections of them, also with confidence intervals for the 95% levels (Yusuf, et al, 1985, discussed in Petitti 1994, pp. 100-102). Several points are worth making about the results. There is strong evidence for the effectiveness of therapeutic community treatment apparent from these studies. The odds-ratio measure used indicates that studies below one have a positive effect, those above one a negative effect, and those on or about one are neutral. However it is vital to consider the confidence intervals for each study, to ascertain that the odds-ratio was unlikely to have happened by chance This is conventionally expressed through the calculation of the range over which the result would be unlikely to have happened more than 5 times out of a 100 (the 95% confidence interval). 19 of the 29 studies indicated a positive effect, within the 95% level of confidence. The remaining 10 studies all had confidence intervals which straddled the neutral score, of which 8 produced odds ratios above one.

When summary odds-ratios are calculated across all 29 studies, as is the convention with meta- analyses, the strength of this finding is underlined. With a summary odds-ratio of 0.57, and an upper 95% confidence interval of 0.61, this set of studies gives very strong support to the effectiveness of therapeutic community treatment. A check can be made on this by grouping the studies. Odds-ratios calculated separately for the RCTs, and for the democratic, concept, and secure types of communities all show strong results, with upper confidence intervals well below one. It is important to note that the RCTs were scattered across the different types of community. This suggests that there was no one subset of studies that was strongly affecting the overall summary result:

Conventionally, this meta-analytic data is presented graphically, as follows. An odds-ratio between zero and one indicates some positive effect, around one indicates a neutral effect, and above one indicates a negative effect. The overall sum, 0.567, is marked with the dotted line:

Publication bias - a funnel plot

Considerable efforts have been made to track down unpublished or "grey" material for this report. In contrast to our expectations, very little turned up. Of concern in meta-analysis is the possibility that publication bias might occur as a result of negative or neutral findings being either nor submitted, or not accepted by journals. A check on any publication bias is provided by lotting the meta-analysis odds-ratios against sample size in a "funnel plot". The lower the sample size, the higher should be the range of odds-ratios reported, giving rise to a typical funnel shaped scattergram. The expectation is that a scattergram would reveal blank spots caused by unpublished findings, or "lost" studies. The funnel plot for this meta-analysis does not suggest that this is the case:

Methodological issues

Evaluating the effectiveness of therapeutic communities, towards the establishment of which this review has been primarily aimed, depends crucially on a clear understanding of what the therapeutic community is, the setting, whether secure or non-secure, in which it is delivered, and at which patients it is aimed. These elements are all evidently contestable, both within a largely sympathetic literature, and within a smaller, hostile literature.

British Mental Health Law has been ambivalent about psychopaths. Both the 1959 and 1983 Acts separate psychopathy from other conditions, and define it behaviourally, while holding a pessimistic view of treatment interventions. Gunderson,1994, p.12, suggests that "personality disorder is a diagnosis whose construct (i.e. its meaning) has grown rapidly and changed dramatically during the past 25 years". Recent official reports on work in this area have reflected these difficulties. For example the Reed Report on Psychopathic Disorder, 1994, states that "the diverse meanings attached to psychopathic disorder often undermined the effectiveness of evaluation of treatment" (p.34). Conceptual expansion makes judgements about research reports difficult, especially if they are more than about fifteen years old.

The definition of therapeutic communities has also been difficult. There are two main types of therapeutic communities: democratic and concept-based/hierarchical. For some writers these are variations on a basically common theme (Sugarman, 1984) - one dealing with deeper intrapsychic change and the other with initial behavioural control; for others they have nothing in common but the name (Glaser, 1983). They have emerged from quite separate origins. In general the intensity, or dosage, of treatment is commonly recognised in the literature by differentiating between therapeutic community approaches and the therapeutic community proper. The former refers to a therapeutic approach across whole hospitals, whereas the latter refers to specialised therapeutic communities dealing with a defined population. In addition, democratic type therapeutic communities developed in prisons or secure settings are inevitably influenced the requirements of prison regulations concerning security and control.

The methodological issues arising from the studies reviewed are numerous. In the 1994 Cochrane Lecture, McPherson, 1994, pointed out that RCTs are important where there is obvious uncertainty, but that they should not be used where there are ethical problems, a lack of objective outcome measures, resistance from the field, or a reluctance to compare treatments. On these grounds we do not feel that there is any intrinsic reason why RCTs should not be mounted further for therapeutic communities. Why have so few well-designed studies been done? The ideal of an RCT has generated difficulties where it has been attempted. Perhaps the most famous attempt was the Clarke & Cornish, 1972, study undertaken over 25 years ago, but which, in the end, proved impractical, and generated a methodological alternative, the cross-institutional design. Some of the difficulties reported in the literature include treatment complexity, treatment dosage and treatment integrity, population selection, dropouts, effects decay, and diagnostic shift.

On the basis of the positive meta-analysis results, it is suggested that in addition to further RCTs, a more complex cross-institutional study is undertaken, together with further cost-offset studies to complement those few already developed.

Discussion and research recommendations

This systematic international literature review has led us to conclude that therapeutic communities have not produced the amount or quality of research literature that we might have expected, given the length of time they have been in existence, and the quality of staff we know exists and has existed in therapeutic communities. This may be partly due to a lack of emphasis placed on research in the early days of therapeutic community development, and more recently to a lack of resources, in terms of finance, staff and adequate research methodologies, designs and instruments. However, it is clear that since the meta-analysis indicates that existing research is in favour of therapeutic communities, there should be more, and more good quality, and comparative, research on therapeutic communities, in order to confirm the case that therapeutic communities are effective, especially since they are expensive. In addition there is clinical evidence that therapeutic communities produce changes in people's mental health and functioning, but this needs to be further complemented by good quality qualitative and quantitative research studies.

Recommendation one There is meta-analytical and clinical evidence that therapeutic communities produce changes in people's mental health and functioning, but this needs to be further complemented by good quality qualitative and quantitative research studies.

There is accumulating evidence, albeit it at a low level of research, of the effectiveness and particular suitability of the therapeutic community model to the treatment of personality disorder, and particularly severe personality disorder. In the absence of conclusive evidence of the effectiveness of any alternative treatment we ought to protect and develop those therapies which can demonstrate some efficacy in treating personality disorder.

Recommendation two Further research on the effectiveness of therapeutic communities for personality disorders is warranted.

There is also evidence of the efficacy of therapeutic communities, modified for prison security needs, in managing difficult prisoners, and significantly reducing serious prison discipline incidents after admission, including fire setting, violence, self-harm and absconding. The placement of a therapeutic community within a secure environment however poses some problems. There are often conflicts between the need to maintain security and control (which is regarded as the primary task of prisons) and the provision of therapeutic community treatment, since therapeutic communities ideally devolve major decisions regarding organisation, rules, treatment, sanctions, admission and discharge, to its clients.

Recommendation three The development of modified therapeutic communities in prisons in the USA and Germany has grown rapidly. The efficacy of this approach should be considered for a research-based demonstration programme in the UK.

There is evidence that the longer a resident stays in treatment, the better the outcome. Very short stay residents do particularly badly.

Recommendation four While a few communities such as Winterbourne Day Hospital in Reading, are tackling this issue, small research projects should be mounted to identify ways of reducing drop-out rates.

Why have so few well-designed studies been done? The ideal of an RCT has generated difficulties where it has been attempted. Perhaps the most famous attempt was the Clarke & Cornish, 1972, study undertaken over 25 years ago, but which in the end proved impractical, and generated a methodological alternative, the cross-institutional design.

Recommendation five A cross institutional design for a study of therapeutic communities "in the field" should be undertaken.

Concept-based therapeutic community research literature is quite considerable, although of varying quality. However, there is sufficient literature around to warrant a literature review and meta-analysis in its own right.

Recommendation six A. review of concept-based therapeutic community literature should be commissioned to complement the current review, with a concomitant meta- analysis based on the studies found.

Globally, the modified therapeutic community seems to be surviving, and proliferating best (especially in the USA) in prisons, and for substance abusers; and concept-based therapeutic communities appear to predominate, both in terms of numbers of therapeutic communities, and in amount of literature, and research generated - although again much of it is of variable quality and generalisability. Concept-based therapeutic communities have also exercised themselves much more than democratic therapeutic communities about the reasons and prevention of early drop- outs, or "splittees". The health service in the main, and particularly in Britain, seems to have neglected the therapeutic community form of treatment, although there appears to be a resurgence of interest recently.

References

Clarke, R.V.G. & Cornish, D.B., 1972, The Controlled Trial in Institutional Research - Paradigm or Pitfall for Penal evaluators, London, HMSO.

Glaser, A.N., 1983, Therapeutic Communities and Therapeutic Communities: A personal perspective,International Journal of Therapeutic Communities , 4, 2, 150-162

Gunderson, J, 1994, Building structure for the borderline construct, Acta Psychiatrica Scandinavica, 89, (Suppl.379),12-18

Kahn, H.A. & Sempos, C.T., 1989, Statistical Methods in Epidemiology, Oxford, Oxford University Press

McPherson, K., 1994, The best and the enemy of the good: randomised controlled, uncertainty, and the role of patient choice in medical decision making. The Cochrane Lecture, Journal of Epidemiology and Community Health, 48, 6-15

Petitti D.B., 1994, Meta-Analysis, Decision Analysis, and Cost-Effectiveness Analysis, Oxford, Oxford University Press

Rapoport, R.N., 1960, Community as Doctor. New Perspectives on a Therapeutic Community, London, Tavistock Publications

Reed Report, 1994, Report of the Department of Health and Home Office Working Group on Psychopathic Disorder, D.O.H./H.O., Chair - Dr. John Reed

Roberts, J, 1997, How to recognise a Therapeutic Community, Prison Service Journal, 111,4-7

Sugarman, B, 1984, Towards a new, common model of the therapeutic community: Structural components, learning processes and outcomes, International Journal of Therapeutic Communities, 5, 2, 77-98

Association of Therapeutic Communities Home Page