Using discrete choice modelling in priority setting: an application to clinical service developments
Introduction
All countries face decisions regarding the level of resources to allocate to the provision of health care and how to allocate these resources within that sector (Ham, 1997). For countries with a publicly-financed health care system, these decisions are not governed by consumer sovereignty or individual choice. In the absence of market forces as the basis of resource allocation, policy makers seek alternative means of choosing between competing demands on the health care budget. In other words, they set priorities. This priority setting activity has always been a feature of publicly financed health care systems. However, during the last ten years, such decisions have become much more explicit and frameworks to assist the process are still being developed. These include approaches such as league tables of quality-adjusted life years (QALYs), needs assessment and programme budgeting and marginal analysis (Mooney et al., 1992, Donaldson and Ratcliffe, 1995, Stevens and Raftery, 1997).
More recently, debate has turned to the issue of who should make these resource allocation decisions (Harrison and Hunter, 1994, Ayres, 1996, New, 1996, Lomas, 1997). In the UK, the Department of Health, 1991, Department of Health, 1992 recommend that policy makers involved in setting priorities should seek the views of many groups, including local people, general practitioners and health care planners and providers. Given this directive, four main issues can be identified:
- •
whose views should be incorporated into the priority-setting process?
- •
how should these views be elicited?
- •
how should the views be incorporated into the chosen approach to priority setting?
- •
and how can the views expressed be reconciled with other factors such as evidence-based medicine?
Various techniques are used to elicit such views for the priority setting process. For instance, citizens' juries (Lenaghan et al., 1996), satisfaction studies (Heyden, 1993), opinion polls (Groves, 1993, Bowling et al., 1993), consensus methods such as the Delphi process (McKenna, 1994, Williams and Webb, 1994), the patient-generated index (Ruta et al., 1994), consensus panels (Stronks et al., 1997), focus groups (Bowie et al., 1995) and willingness to pay (WTP) (Olsen and Donaldson, 1998, Donaldson et al., 1997).
Shackley and Ryan (1995) argue that the method of eliciting community values should incorporate the notion of opportunity cost and reflect the strength of preference of individuals. Of the aforementioned approaches, only WTP incorporates both of these properties. However, WTP is often viewed as a controversial means of eliciting values. In this paper we use a related technique known as discrete choice modelling (DCM). This technique addresses the requirements identified by Shackley and Ryan (1995) whilst avoiding the controversy raised by the use of WTP.
This paper explores the use of discrete choice modelling (DCM) to elicit the views of planners of health care, providers of health care and consumers in the area of priority setting. The technique can be used to estimate ‘benefit scores’ and, therefore, produce output that can be used within an economic evaluation framework to assist the process of priority setting. In the background, a specific priority setting problem faced by a health care provider is described — namely, which clinical service developments to choose from the many contesting cases given the limited budget. The provider was concerned to take into account the views of the hospital consultants (senior medical personnel) in the decision-making process.
Following the background, the rationale for using DCM is discussed and the technique described. Empirical work is then presented, showing the application of DCM to assessing the benefits of clinical service developments. It is shown how the technique can be used to estimate cost per unit of benefit for a series of proposed clinical service developments and, thereby, help inform decision-makers about the efficient allocation of scarce health care resources. The results are discussed and it is proposed that integer programming, alongside DCM, is useful for selecting the optimal combination of clinical service developments within a given budget.
Section snippets
Background
Since 1990 health boards across Scotland have been charged with the responsibility of purchasing health care services and more recently (following the release of the Scottish Office Department of Health White Paper) leading the development of Health Improvement Programmes, to meet the needs of the local population (Department of Health, 1997, Scottish Office Health Department, 1998). In both of these roles, the health boards have an annual budget with which to meet their responsibilities. Each
Discrete choice modelling
Discrete choice modelling is a stated preference technique. It has evolved from conjoint analysis, an approach to eliciting values originally developed by mathematical psychologists for applications in market research. The historical development of conjoint analysis and DCM and their application is described in Ryan (1996) and in this journal in Ryan (1999). When applied in a health care context, conjoint analysis and, more specifically, DCM have generally been used to value patient benefits
Methods
There were seven main stages in the study. The first five are typically carried out as part of a DCM or conjoint analysis study2: establishing the dimensions; assigning levels to the dimensions; identifying which scenarios to present to consultants; eliciting consultants' preferences via a questionnaire; and estimating the regression equation to give weights to the
Results
One hundred and thirty of the 216 hospital consultants responded within the required two weeks time scale, giving a response rate of 60%. One hundred and twenty six consultants gave their gender and of these 21 were female. The mean average age of respondents was 46 years, the youngest was 33 and the oldest 67. Seventy per cent of the consultants had some responsibility for clinical management.
The average time to complete the questionnaire was 11.5 minutes, with a minimum of two and a maximum
Discussion
This study had both practical and academic objectives. At the practical level, a secondary health care provider required a system to help set priorities for new clinical service developments. They wanted this system to incorporate the views of hospital consultants. This is the first time, to the authors' knowledge, that discrete choice modelling has been used in a priority setting context. The information available for informing decisions regarding choice of clinical service development
Conclusion
In conclusion, the results of this study provided valuable information which was used by the senior management of the Trust in the prioritisation of clinical service developments. Future work is needed to address the practical issues raised by this study, as well as the more general methodological issues that are raised in the development of a new instrument (Ryan et al., 1996). However, the study suggests that discrete choice modelling has potential as a useful tool in the area of priority
Acknowledgements
We are grateful to Aberdeen Royal Hospital Trust Medical Advisory Committee, especially Dr. J.A.R. Friend and Professor J. Hutchison, for their advice on the construction of the questionnaire and to all the consultants who completed the questionnaires. Thank you also to Joanna Fawcus who helped to administer the questionnaire. We also acknowledge that the project was initiated as a result of discussions by D.R. with Joanna Reid in her capacity as General Manager of the Medical Directorate of
References (52)
Rationing health care: views from general practice
Social Science and Medicine
(1996)- et al.
Applications of cost-benefit analysis to health care. Departures from welfare economic theory
Journal of Health Economics
(1987) - et al.
Health service priorities: explorations in consultation of the public and health professionals on priority setting in an inner London health district
Social Science & Medicine
(1993) Priority setting in health care: learning from international experience
Health Policy
(1997)- et al.
Helicopters, hearts and hips: using willingness to pay to set priorities for public sector health care programmes
Social Science & Medicine
(1998) Using conjoint analysis to go beyond health outcomes: an application to in vitro fertilisation
Social Science & Medicine
(1999)- et al.
What do patients value? Willingness to pay for ultrasound in normal pregnancy
Medical Care
(1985) - et al.
Consulting the public about health service priorities
British Medical Journal
(1995) - Bradley, M., 1991. User's manual for SPEED version 2.1. Stated preference experiment designer. Hague Consulting...
- et al.
Magnetic resonance imaging for the investigation of knee injuries: an investigation of preference
Health Economics
(1998)