Is There Still a Place for Independent Research into Issues of Public Policy in England and Wales in the 1980's? A Case Study from the Field of Health Care: Modelling Hospital Costs
This paper is concerned with the viability of independent research into issues of public policy in a climate which is becoming increasingly hostile. The general argument is illustrated by one specific example relating to the development of models of hospital inpatient costs in the National Health Service. In a recent publication by staff of the Operational Research Service of the Department of Health and Social Security a model (denoted the C-G-N model) of hospital costs is put forward. However, no mention is made of an alternative model (denoted the A-B model) published some two years previously and we have therefore decided to compare the two models in terms of general structure, fit to the available observational data and applicability. It is shown that the C-G-N model was applied by its authors in a form which involves an heteroscedastic residual error term. When this model is fitted to a representative data set, with an appropriate weighting factor which produces homoscedastic errors, substantial differences are found in the corresponding parameter estimates. The C-G-N model in its original and modified versions is then compared with the A-B model, and it is shown that the latter is more parsimonious in terms of parameters, is applicable to a wider range of types of hospital and fits the available empirical data somewhat better than either version of the former. Furthermore, there are substantial differences in the conclusions which might be drawn on the basis of the various models in several important classes of application. An outline is given of a more complex model, denoted the component model, which incorporates features of both the C-G-N and A-B models, provides a better fit to the empirical data than either and has a structure which is more satisfactory for a wide range of applications.