A multivariate analysis of some characteristics of psychiatric hospitals

Abstract

The psychiatric hospitals in England and Wales differ widely in the ways in which they are run and on a range of other
criteria such as staff-patient ratios and discharge rates. Many of these differences have come about for historical reasons. Some patients are treated in the former 'county asylums' set up over a century ago while other patients are treated in psychiatric units in the new district general hospitals. Other differences may be due to the location of the hospital and the make-up of its catchment population.
This wide spread variation was observable on a large selection of data, much of which is collected and published regularly by the Department of Health and Social Security. It was decided that the problem of comparing hospitals with a view to accounting for their differences could be profitably tackled using various multivariate statistical techniques. Principal components analysis and canonical correlations analysis were used to select a small number of important variables from the large number available and a form of causal analysis was used to examine the relationships between this reduced set of variables.
The variables chosen dealt with a psychiatric hospital's four main functions which were seen to be custodial, protective, therapeutic and socialization. A set of 'performance' variables, which covered the therapeutic function, was first chosen, since this was considered to be the most important function. Then several 'explainer' variables, which were divided into sets dealing with environmental, professional, institutional and socio-medical aspects of a hospital's functioning, were chosen. These contained variables dealing with a hospital's remaining three functions and variables thought to explain differences between the variables which dealt with a hospital's functions.
The preliminary analysis identified several types of hospital characterized by their values on certain variables. In particular a kind of hospital which was called the 'revolving- door' type was observed. This had high discharge and turnover rates, few in-patients but many out and day patients. In addition it had high staffing and expenditure, good accessibility, fewer social workers. The wards were uncrowded with low bed-occupancy and some patients were able to work outside the hospital.
The causal analysis hypothesized that the variables were related in a certain order of causation and showed that several of them had only spurious correlations with one another which disappeared when the effect of other variables was removed by partial correlation. The most important feature of the 'revolving-door' hospitals which remained was that variations in the discharge rate were primarily due to variations in staffing ratios and expenditure on certain items.
It can thus be seen that variations in important aspects of a hospital's functions can be at least partially explained by the methods outlined above. The results depend to a large extent on the suppositions incorporated in the causal analysis. Despite this they can be of value to those who may wish to reduce variation between hospitals or to alter various aspects of their functions so that they carry them out in the way which is currently considered most suitable.