We were delighted to read that our paper was so enthusiastically discussed by Sridharan et al at their evidence-based journal club. They have spotted the main limitation to the study, which was included in our own list of limitations — namely, that our findings were “based upon only one service model and may have limited generalisability” (Sipos et al, 2001). In our paper, we cited previous work from Nottingham (Harrison et al, 1991), showing how the development of multi-disciplinary teams had coincided with a reduction in the proportion of patients with first-episode psychosis requiring hospitalisation at initial contact. In Sipos et al (2001) we went on to show that, although there is clearly a reduction in hospitalisation at first contact, the risk of admission at some point in the first 3 years after first onset has actually remained the same. Indeed, there are striking differences between those patients admitted early in the course of the disorder and those admitted later.

On reflection, we agree that the paper would have benefited from a slightly more detailed specification of service changes in Nottingham, although these have been described elsewhere and we would refer readers to Beck et al (1997). We would caution, however, against attempts to draw causal inferences from the presence, or absence, of particular ‘community’ services because our paper reported an observational study rather than a controlled one. The research community has barely begun to understand the interplay between different components of ‘community-oriented’ services and patient outcomes. The parameters mentioned by Sridharan et al are certainly pointers in the right direction but we have some way to go in describing (and measuring) factors such as the amount of ‘social support’ available, let alone evaluating their impact on outcomes.