This review synthesised numerous individual studies on psychosocial interventions for patients diagnosed with schizophrenia. The search was restricted to German and English language papers, which leaves some room for language bias. None of the included studies were unpublished. The authors provided detailed inclusion criteria on the studies, but no information on study selection and data extraction processes, i.e. what steps were undertaken to reduce bias and errors. In addition, they did not report whether they assessed the validity of the studies formally. The review was restricted to studies that provided specific data.
The control groups varied widely in the treatment offered (i.e. only ordinary clinic treatment, occupational therapy, assisting psychotherapy). This meant for some studies that two psychosocial interventions were compared: one unspecific (i.e. assisting psychotherapy) and one specific (i.e. psychoanalysis).
The outcomes assessed in the individual studies varied greatly (e.g. social behaviour measures, relapse rates), not much information about the assessment method was given, and all were combined in the effect size measure.
The random assignment to treatment and control groups was not part of the description of the included studies, nor was it an eligibility criterion for inclusion in the review. This weakens the evidence base, as it is unclear whether the control and treatment groups in the individual studies differed already before the start of the treatment. A separate analysis on randomised controlled trials only was not reported.
Heterogeneity between the studies was ignored in favour of determining an overall effect size for psychosocial interventions. The effects of individual therapies and different follow-up periods were also presented.
The mean effect size was translated into other measures to provide further explanation.
The authors' conclusions are plausible, but caution is required in view of the evidence base.