As the authors of a previous review of deinstitutionalisation and homelessness, ^{Reference Craig and Timms1} we were interested to see our 1992 findings confirmed by Winkler et al's recent paper ^{Reference Winkler, Barrett, McCrone, Csémy, Janousková and Höschl2} and the accompanying editorial. ^{Reference Salisbury and Thornicroft3} We particularly agree with the notion that apparent relationships between deinstitutionalisation and homelessness can often be mediated by substantial confounding factors. In London in the 1980s, it was the unheralded and unpublicised closure of most of the city's homeless hostel beds that seemed the most likely culprit.

Although this issue may well still be pertinent in other healthcare and social systems, it was of decreasing relevance in the UK even when we published our paper in 1992. The process of deinstitutionalisation was, by then, irreversible and substantially accomplished. This leads us to our concern that these papers might support an unhelpful sense of complacency.

Taylor Salisbury & Thornicroft's statement that ‘instances of homelessness … among those discharged are rare’ ^{Reference Salisbury and Thornicroft3} is clearly correct in referring to the institutional closures and hospital discharges that are now several decades in the past. However, it is at odds with the situation of hospital discharge as it stands today, at least in London. I work in a psychiatric outreach team for homeless people in South London, where homelessness following hospital discharge is common among referrals to our service. We looked at 3 months of our referral data last year and found that 60% of our homeless referrals (mainly with a diagnosis of psychosis) had had previous contact with our local mental health service. They had had, on average:

1. • contacts with 4 separate trust services

2. • 35 contacts (face-to-face/phone triage), 2 of these would have been emergency contacts, seen in an accident and emergency department, or in a section 136 suite

3. • 65 days as an in-patient in the local trust service.

These people had sometimes been discharged to the street, or referred to local community services, but without effective plans to prevent them becoming homeless again. We note that observations we made in 1992 still stand – the excessive bed occupancy of in-patient services is driving an emphasis on short episodes of in-patient treatment.

It seems clear that a small but significant number of people are simply ill-served by the existing format of mainstream mental health services. It may be (as I have heard in a European ‘quality’ forum) that such people are just peculiarly difficult. This seems unlikely, given a recent outcomes study we did of the most alienated and intractable of our referrals – people who live on the street and who have not been engaged by the sustained efforts of experienced street outreach teams. The intervention concerned was involuntary admission to hospital under a section of the Mental Health Act. ^{Reference Timms and Perry4} One year later, the majority were still engaged with the specialist mental health team and were still in accommodation. Here is an area ripe for research – the vital factors that enable such teams to engage effectively, and to maintain that engagement, with homeless people with psychotic disorders.