Skip navigation

Preventing Homelessness


Randomised controlled trials of social interventions are rare, so Bandolier was delighted to spot a well-conducted trial looking to prevent recurrent homelessness in mentally ill men in New York [1]. This is a detailed paper with a wealth of information about its conduct, so for those who are interested, then it comes into Bandolier's "must-read" category.

Subjects

The study population was 102 men discharged to housing in New York City region in 1991-93. Of these, 96 agreed to participate in the trial. All the men had severe mental illness, usually schizophrenia or other psychotic disorders. They were discharged from an on-site psychiatry programme in a mens' shelter, to return to community housing. They had access to a broad spectrum of supportive housing, from intensively supervised residences to single-room-occupancy hotels with on-site social services. But discharge to family, or friends or other arrangements was common.

Intervention

Randomisation was between usual services only (USO), and a critical time intervention (CTI). The critical time was defined as the first months after discharge, and the intervention included a range of services (Table) provided by a CTI worker (no special skills, though supervised by a mental health professional). The CTI workers were "street smart", and gave as much as was needed by individual patients.
Services received by mentally ill men discharged to the community
Months after discharge CTI group USO group
CTI workers: Shelter staff:
1-3 Make home visits Assist patients and caregivers on request
Accompany patients to appointments Substitute for caregivers when necessary
Meet with caregivers
Substitute for caregivers when necessary
Give support and advice to patient and caregivers
Mediate conflicts
Negotiate ground rules for relationships
4-7 Observe trial of ground rules Services provided by community
Help modify ground rules as necessary Phone advice for patient or caregiver
8-9 Reaffirm ground rules
Hold parties/meetings to symbolise transition
10-18 Usual services

Outcome

The main outcome was the number of homeless nights (not including nights when patients decided not to go home because they had other things to do). This was assessed by monthly face-to-face assessments conducted by assessors blind to the intervention. This was continued monthly for 18 months.

Results

Two men, both in the USO group, were lost to follow up, one who was fleeing drug dealers and one who committed suicide on becoming HIV positive. There were no differences between the groups, and cocaine and alcohol abuse was high (about 50%).

Overall the 48 men in the USO group had 4370 homeless nights, compared with 1415 in the CTI group. Over 18 months (548 nights), the average number of homeless nights was higher at 91 in the USO group than the 30 nights of the CTI group. Fewer men had extended periods of homelessness with the critical time intervention (Figure).



During the last month of the 18-month follow-up, only 4 men in the CTI group were homeless, compared with 11 in the USO group - relative risk 0.36 (0.12 to 1.06), NNT 6.9 (3.5 to 284). Extended homelessness (more than 54 nights) occurred in 10 men in the CTI group and 19 in the USO group - relative risk 0.53 (0.27 to 1.01), NNT 5.3 (2.7 to 130).

Comment

This was a small trial in terms of numbers, but a big trial in terms of care taken during a prolonged follow-up. It demonstrates a strategy shown to be effective - preventing one case of homelessness for every five receiving the intervention. It was directed specifically in the prevention of homelessness in mentally ill men, in social conditions likely to be much worse than those prevalent in the UK. Those involved in healthcare and social services - in provision of services and in policy-making - would profit from reading this paper.

Reference:

  1. Preventing recurrent homelessness among mentally ill men: a "critical time" intervention after discharge from a shelter. American Journal of Public Health 1997 87: 256-62.



previous or next story in this issue