The British Journal of Psychiatry
Effects of training community staff in interventions for substance misuse in dual diagnosis patients with psychosis (COMO study): Cluster randomised trial
Cluster randomised trial


A cluster randomised controlled trial was used to investigate the effectiveness of training staff in 13 London community mental health teams (CMHTs) to deliver substance misuse interventions to patients with psychosis and comorbid substance misuse (‘dual diagnosis’). The primary hypotheses, which were that experimental group patients would spend fewer daysin hospital over 18 months of follow-up and show reduced alcohol and drug consumption, were not confirmed, although confidence intervals were wide for some outcomes. Current UK policy guidance advocates training CMHT professionals to deliver dual diagnosis interventions, but the effectiveness of this strategy has not so far been demonstrated.

The adverse outcomes associated with psychosis and comorbid substance misuse (‘dual diagnosis’) have been well documented, but treatments supported by substantial evidence are few (Jeffery et al, 2000; Tyrer & Weaver, 2004). Interventions that show some benefit have tended to involve relatively intensive treatment of selected populations by specialist therapists (Barrowclough et al, 2001; Bellack et al, 2006); generalisability to routine settings is unknown. However, commentaries and policy guidance on dual diagnosis in England have favoured ‘ mainstreaming’ dual diagnosis interventions by integrating them into care provided by existing clinical teams (Johnson, 1997; Weaver et al, 1999; Department of Health, 2002). The effectiveness of this strategy is untested.

Our aim was to investigate whether a training and supervision intervention delivered to community mental health team (CMHT) case managers would improve patient outcomes. At the patient level, the primary hypotheses were that, compared with controls, patients on the case-loads of experimental group case managers would make less use of in-patient services and would be consuming smaller quantities of substances when assessed 18 months later.


A cluster randomised controlled trial design was employed, each cluster consisting of the patients on a particular staff member’s case-load.

All permanent case managers in 13 London CMHTs were invited to participate. Their case-loads were screened for patients who met study criteria for dual diagnosis, and all who did were included in the sample. This screening stage involved first identifying patients with clinical diagnoses of schizophrenia, another non-affective functional psychosis, or bipolar affective disorder. With guidance from researchers, case managers rated each of these patients using the Clinician Alcohol and Drug Use Scales (Drake et al, 1996). Patients identified as misusing or dependent on at least one substance met our study criteria for dual diagnosis. Case managers were randomised to intervention or control groups by an independent statistician. All patients identified as eligible for the trial entered the experimental or control group according to their case manager’s assignment.

The experimental intervention consisted of a treatment manual, a 5-day training course in assessment and management of dual diagnosis, and subsequent monthly supervision. Motivational interviewing was a central source (Swanson et al, 1999), and the training also drew on cognitive–behavioural relapse prevention techniques (Irvin et al, 1999). The control group received CMHT management as usual with no specific dual diagnosis intervention. To reduce contamination, experimental group staff were asked to avoid sharing manuals and details of training.

At baseline, socio-demographic and clinical details of all patients were recorded. At baseline and after 18 months, data were collected on the two primary outcome measures: (a) hospital bed use over the preceding 18 months, recorded using best available information from patient interview, clinical records and local electronic patient data systems; (b) substance use over the preceding month, documented at patient interview using the Maudsley Addictions Profile (Marsden et al, 1998).

Secondary outcomes relating to adverse events, symptoms and social functioning and staff-level outcomes were also assessed, but are not reported here. Interviews with patients were carried out whenever possible; for patients who were not available, ethical approval was obtained to gather data from staff on their characteristics and the bed use outcome.


Seventy-nine case managers participated. Of the 1560 patients on their case-loads, 232 met criteria for dual diagnosis. Forty of the 79 case managers were randomised to the experimental group and 39 to the control group. This yielded 127 patients with dual diagnosis on case-loads of case managers in the experimental group and 105 on control group case-loads. Miles et al (2003) have described the characteristics of the sample. Experimental and control groups were similar except for an imbalance for White ethnic group (61% of the control group v. 43% of the intervention group). CONSORT diagrams of staff and patient flows through the study are given in data supplement 1 to the online version of this paper.

Three patients died during the 18-month follow-up period. Of the remaining 229 patients, 77 (62%) of the intervention group and 77 of the control group (74%) were interviewed at follow-up (P=0.079). Bed use data were obtained for 113 intervention and 97 control group members. We defined experimental group patients as having received the intervention as intended if their case managers had attended at least 4 days of training and if they had remained on the case-load of a trained case manager for at least 9 months: just 45 of the 127 experimental group patients met these criteria. Eighty-six of the 105 control group members fitted the study definition of having remained in their intended treatment group, which required them to have remained on a CMHT case-load for at least 9 months and not to have been taken on by a trained case manager.

Details of outcomes are shown in data supplement 2 to the online version of this paper. For bed use, there was no evidence of a difference between experimental and control groups (mean bed use for experimental group: 74.9 days (s.d.=142.6) over 18 months follow-up; for control group 71.8 days (s.d.=128.1), P=0.30 following log transformation and adjustment for baseline). However, standard deviations were higher than anticipated when carrying out the study power calculations, resulting in wide 95% confidence intervals. There was no significant difference in proportion of patients admitted during the follow-up period (43% of the experimental group v. 48% of the control group, P=0.18).

Self-reported alcohol and drug use of interviewed members of each group over the 30-day period before the follow-up interview are also shown in online data supplement 2. Neither the proportion who had consumed substances (74% of the experimental group and 71% of the control group for alcohol, 32% and 36% respectively for cannabis and 16% and 18% for other drugs) nor the quantity consumed over the month differed between groups. No difference in outcomes became significant after adjusting for baseline values.


The study’s strengths lie in external validity: the intervention took place in a routine National Health Service setting and was brief enough to be replicable in such settings, and all identified patients with dual diagnosis were included. Limitations include high attrition from the intended intervention, reliance on clinician substance misuse diagnosis and lack of masking. Fidelity was not measured, and we do not know to what extent case managers implemented the intervention as intended. Also, for the main outcomes, standard deviations were wider than anticipated when power was calculated: confidence intervals are thus wide and include the possibility of a substantial effect in either direction.

There was no evidence that the training intervention affected bed use or substance use. The limitations discussed above must be considered in interpreting this finding. Also, there is much evidence that influencing outcomes among people with dual diagnosis is difficult, with many reported negative findings (Tyrer & Weaver, 2004). Our intervention was of low intensity, and the limited amount of training provided might have been insufficient to influence clinical practice. Our findings thus fail to lend any clear support to the current UK policy of ‘mainstreaming’ dual diagnosis interventions by training staff within generic mental health services to deliver them. Other models for introduction of dual diagnosis interventions into routine clinical settings may therefore need to be tested, taking into account the few available positive findings from efficacy studies in more selected groups (Barrowclough et al, 2001; James et al, 2004). Possible options include specialist dual diagnosis teams and specialist workers within CMHTs. Providing interventions at an early stage of illness when adaptive and maladaptive ways of coping with illness are less well established has so far been evaluated only in small pilot studies (Kavanagh et al, 2004; Edwards et al, 2006). Until further evidence is available about the effectiveness of implementing these models in routine settings, evidence-based policy making in the area of dual diagnosis poses great difficulties.

  • Received January 29, 2006.
  • Revision received April 12, 2007.
  • Accepted April 19, 2007.


View Abstract