A review of surveys of mental hospital populations in Britain in the 1960s carried out for planning purposes reveals conflicting conclusions about the likely requirement for mental hospital beds over the next 15 to 20 years-the period over which estimates are needed for planning facilities. The (Baldwin, 1963;Rehin and Martin, 1963), and the low expectation of separation from hospital by death or discharge of the high proportion of relatively young long-stay patients (Gore and Jones, 1961). Earlier optimism about the possibilities of rehabilitation of long-stay patients and reduction of requirements for long-term beds (Cooper and Early, 1961; Norton, 1961) has been tempered by more recent studies which have expressed doubt as to the likelihood of fulfilment of the original estimates (Hassall and Hellon, 1964;Early and Magnus, 1966). It has also been shown that large differences exist in the behaviour of individual hospitals and of age and sex groups within hospitals, so that a uniform measure of bed needs could be misleading (Hassall, Spencer, and Cross, 1965).The problem of projecting the future size and structure of a mental hospital population on the basis of past and current use is usually subdivided into consideration of two interrelated processes:(1) The expected trend in the number of future admissions and their duration of stay or bed usage.(2) The expected attrition of the standing mental hospital population existing at the beginning of the projection period by death or discharge. This paper deals exclusively with the second of these processes.Tooth and Brooke (1961) based their calculation on the annual rate of separation (i.e. discharge and death) of the standing long-stay hospital population (over 2 years) at the end of 1954, assuming a linear fall to extinction in 16 years. It was subsequently pointed out that the assumption of linear attrition was almost certainly incorrect even on the basis of the original data (Lindsay, 1962). Lindsay showed that application of the annual separation rates to the patients in the original standing long-stay population remaining in hospital at the end of each year would represent a decreasing geometric series and not an arithmetic series, so that projection would result in a curve yielding a quarter of the patients remaining after 20 years, i.e. by 1974 . Norton (1961 had also projected in a curvilinear fashion from data on actual separation experience over 6 years.In order to improve the accuracy and usefulness of projections of this kind, a detailed study was undertaken to develop a method of forecasting which would make use of direct estimates of the probability of separation occuring after specified lengths of stay in hospital for any defined group of patients, such as an age, sex, or diagnostic group. A further objective of the study was to allow for the effect of ageing of the population remaining in hospital, so that the age composition of the residue could also be anticipated for any projection period.
METHODPerhaps the most obvious approach to this prob...