Background. Concerns have been raised about the scope and generalizability of much community mental health research. In particular, both experimental and control services are poorly characterized.Methods. To review the effectiveness of ‘home treatment’ for mental health problems in terms of hospitalization, we conducted a systematic review, using Cochrane methodology but with a wider remit. Non-randomized studies were included in response to concerns about RCTs’ generalizability. All authors were followed up for data on service components. ‘Home treatment’ was defined broadly for the purposes of the literature search, but included studies were then assessed against service components specifically focused on delivering treatment at home. The study tested components and other features for associations with days in hospital, as well as conducting a conventional meta-analysis of data on days in hospital.Results. We found 91 studies, 18 comparing home to in-patient treatment. Sixty per cent of authors responded to follow-up. The vast majority of the services studied had a ‘home treatment function’ and regularly visited patients at home. The heterogeneity of control services made meta-analysis problematical as did the limited availability of data. There was some evidence that ‘regular’ home visiting and combined responsibility for health and social care were associated with reduced hospitalization. The inclusion of non-randomized studies rarely affected the findings.Conclusions. Evidence concerning the effectiveness of home treatment remains inconclusive. A centrally coordinated research strategy is recommended, with attention to study design. Experimental and control service components should be prospectively recorded and reported to enable meaningful analysis.