Objectives -To use routinely collected data to provide a reliable estimate of the size and psychiatric morbidity of the homeless population of a given geographical area by using capture-recapture analysis.
Design -A multiple sample, log-linear capture-recapture method was applied to a defined area of central London during 6 months. The method calculates the total homeless population from the sum of the population actually observed and an estimate of the unobserved population. Data were collected from local agencies used by homeless people.
Subjects -Homeless people in north east Westminster residing in bed and breakfast accommodation and hostels or sleeping rough who had contacted statutory or voluntary agencies in the area.
Results -2,150 contacts by 1,640 homeless people were recorded. The estimated unobserved population was 3,293, giving a total homeless population for the period of around 5,000 (SD 1250). Mental health problems were significantly less prominent in the unobserved compared with the observed population (23% (754) v 40% (627), P<0.0001). For both groups the prevalence varied greatly with age and sex.
Conclusions -Capture-recapture techniques can overcome problems of ascertainment in estimating populations of homeless and homeless mentally ill people. Prevalences of mental illness derived from surveys that do not correct for ascertainment are likely to be falsely inflated while at the same time underestimating the total size of the homeless mentally ill population. Population estimates derived from capture-recapture techniques may usefully provide a good basis for including homeless populations in capitation calculations for allocating funds within health services.