This thesis addresses the problem of controlling hospital costs and seeks to provide the basis for a partial solution. Hospital funding in Australia and overseas is known to be deficient in the area of rewarding the more efficient hospital at the expense of the less efficient. Containing hospital cost increases within prescribed limits, as determined by health authorities or other third-party payors, does not necessarily promote the goal of more effective and efficient use of resources. It is now widely accepted that hospital resource planning and administration can be improved through the application of practices based on appropriate measures of hospital output. Once an acceptable measure of output is developed and appropriate unit costs are determined for the different products, then control of hospital costs will be greatly enhanced.
Acceptable measurement of hospital output hinges upon an appropriately and correctly specified classification of the different types of products produced. Given the major part of the hospital's work is directed towards inpatient treatments, a classification of the different types of cases treated or patient case-mix is required. Various methods have been applied to the problem of hospital output definition and measurement over recent decades. Some outstanding advances have been made in the United States with the most notable system of patient case-mix classification, known as diagnosis-related groups, being adopted as the basis for prospective reimbursement of hospitals. Despite this application the system has not yet been universally accepted by the hospital and health insurance sectors. Although much criticism has understandably been levelled at the inadequacies of the diagnosis-related groups (DRGs) or the vagaries of the new reimbursement mechanism, the most telling criticism has gone largely unnoticed.
The DRG scheme and others like it simply lack sound economic theory. In particular, where the classifications are constructed using large data bases, there has been little attempt to ensure that the principles and procedures of numerical taxonomy are properly linked to an economic theory of the underlying cause which the classification seeks to explain. The problem is not a new one as Shubik(1970) has noted that numerical measurement in economics has been fraught with certain difficulties. Shubik(1970) predicted that the principles of and procedures of numerical taxonomy would begin to be used much more in economics, particularly in areas such as consumer and demand theory where measurements of economic phenomena are required. However, despite Shubik's prediction, there has been little application of numerical taxonomy in economics, except for notable efforts by researchers such as Sonquist and Morgan (1964), No apparent effort has been made to use numerical taxonomy in combination with economic tools to develop new theories of economic behaviour of consumers or firms. This thesis provides a useful advance to our body of knowledge in this area.
A theory is developed in this thesis using the characteristics approach, proposed by Lancaster(1971), to explain the behaviour of hospitals. Hospitals are viewed as firms or organisations which produce a number of different outputs, such as patient treatments, education and research. Since a large part of the output of hospitals is patient treatments, the theory concentrates on this aspect of production. Patient treatments are characterised by certain endemic features which enter into the decision functions of both the patient-consumer and hospital-provider. These features are incorporated into a formal description of clinical decision-making in terms of resource use as a further contribution to the construction of a sound classification of patients. Hospital resource use and production are thus explained in terms of these characteristics of patient treatment. Optimal resource allocation and product mix can be determined using this theory.
A further contribution is made by linking and extending the economic model of hospital production to patient case-mix classification. The thesis shows that as certain features characterise particular products, then separate products can be identified by their possession of specific combinations of characteristics. This phenomenon is useful for classifying patient treatments into product groupings useful for effective and efficient management. The thesis describes the analysis of patient data and development of a sound, valid and acceptable patient case-mix classification which incorporates appropriate elements of economics, medicine and numerical taxonomy.