Elsevier

Journal of Health Economics

Volume 25, Issue 5, September 2006, Pages 927-944
Journal of Health Economics

Using discrete choice experiments to estimate a preference-based measure of outcome—An application to social care for older people

https://doi.org/10.1016/j.jhealeco.2006.01.001Get rights and content

Abstract

This paper reports the results of a study that used discrete choice experiment (DCE) methodology to estimate quality weights for a of social care outcome measure. To reflect different states of need, a five-dimensional profile measure was developed. Experimental design techniques were used to derive a sample of states for which preferences were elicited. The DCE approach was employed to elicit values and regression techniques used to estimate a model that could predict index scores for all 243 possible outcomes. The utility model, estimated on the basis of 297 responses, had good test–retest reliability and converged with preferences elicited from a rating exercise.

Introduction

One of the greatest challenges facing health economists is the identification and valuation of benefits from health care interventions. Up until the 1990s, benefit assessment in health economics was dominated by the assumption that health was the only important outcome from health care. This is evidenced by the large amount of research devoted to valuing health outcomes using quality adjusted life years (QALYs). The 1990s however saw a challenge to this assumption. It was recognized by some health economists that concentration on health outcome fails to allow for the possibility that individuals derive benefit from what were being called non-health outcomes and process attributes (Mooney and Lange, 1993, Mooney, 1994, Ryan and Shackley, 1995, Ryan and Hughes, 1997, Donaldson and Shackley, 1997, Donaldson and Shackley, 2003, Donaldson et al., 1997, Donaldson et al., 1998, Protière et al., 2004). Discrete choice experiments (DCEs)1 were developed within health economics to allow inclusion of more than just health outcomes (Ryan and Gerard, 2003), with more recent applications going beyond those included in generic QALYs (Sculpher et al., 2004). The study presented here examines the use of the DCE approach to estimate quality weights within the QALY framework.

The two approaches currently preferred by economists to estimate such weights are standard gamble (SG) and time-trade off (TTO) (Dolan, 1997, Brazier et al., 2002). There is evidence that, when individuals are asked to trade between attributes, what becomes salient is the specific attribute that is involved in the trade (Maas and Stalpers, 1992). This has been shown within SG experiments. Here individuals first are asked their probability indifference point between a gamble and certain outcome. They are then asked to make an actual choice between these same two scenarios (i.e. the certainty equivalent and the gamble to which individuals said they were indifferent). The certainty equivalent is then usually preferred (Von Winterfeldt, 1980, Tversky et al., 1988). There is thus a disjuncture between the stated preference and the actual choice. The same sort of problem may well arise with the TTO technique. DCEs potentially overcome this so-called salience problem since individuals are asked to consider a number of attributes at once.

The policy context for the application of DCEs to estimate QALY weights is older people's preferences for the outcomes from social services. Such services are concerned with managing the effect of (usually chronic) impairment on people's daily lives. Health-based outcome measures that pick up changes in functions or ability are often inappropriate for social services. Moreover, the degree to which generic QALYs can be used with respect to the needs of older people has been questioned (Donaldson et al., 1988). These researchers showed that, in the field of evaluating care for older people, as compared with programme-specific methods, generic quality of life measures are less sensitive to changes in people's health states. Generic health outcome measures, such as EQ-5D and SF-6D (Dolan, 1997, Brazier et al., 2002) are therefore unlikely to be adequately sensitive. Despite that paper being published 17 years ago, it remains the case that a programme-specific utility measure for older people has not yet been developed. Such a development is an important contribution of this paper. While the results presented here provide some insight into the value older people place on different outcomes of social care, nonetheless the primary objective is to demonstrate the feasibility of using DCEs to estimate quality weights.

We start by describing the DCE approach to valuation. We then discuss the methods and results of a DCE study that devised a preference-based measure of outcomes for social care of older people. There were three main components to this study: (i) developing a profile instrument for social care for older people, with appropriate domains and levels; (ii) estimating preference-based quality weights for this instrument, resulting in the older persons’ utility scale (OPUS) and (iii) testing the reliability and validity of the OPUS instrument and quality weights. In this paper we focus on the preference-based quality weights, as well as the reliability and validity of these weights. (For more information on the development of the profile measure, as well as the reliability and validity of this profile measure, see Netten et al., 2002.)

Section snippets

Discrete choice experiments

Discrete choice experiments (DCEs) have been used widely to elicit values in a number of areas, including market, transport and environmental economics (Louviere et al., 2000). The last 15 years have seen an increasing use of the technique in health economics (Ryan and Gerard, 2003). While a limited number of published studies have adopted the DCE methodology to estimate values for different health state profile (Hakim and Pathak, 1999, Johnson et al., 2000, McKenzie et al., 2001), none to date

Deriving the profile instrument for social care for older people

The principal outcome of social service interventions is the management of the effect of long-term impairment, usually in terms of meeting needs for help and assistance. Two comprehensive pre-existing measures of need were taken as a starting point: the Camberwell Assessment of Need (CAN) which provides a comprehensive assessment of the clinical and social needs of people with mental illness (Phelan et al., 1995) and the Camberwell Assessment of Need for the Elderly (CANE) (Orrell et al., 1997

Results

Of the 357 subjects interviewed for the main preference elicitation experiment, 326 passed the consistency checks and completed all items of the questionnaire (21 individuals ‘failed’ two consistency tests, and 10 did not complete all items, with one of these being a respondent who also failed the consistency checks). Table 2 provides details of this main sample and the number of individuals resampled for use in subsequent analysis. Compared with people over 60 in the general population,

Discussion and conclusion

This paper has advocated the use of the DCE methodology to estimate quality weights within the QALY framework. It was shown that the resulting regression equation could be used to estimate utility weights within the framework of a programme-specific QALY measure. An older person's utility scale (OPUS) was developed, which showed good test–retest reliability and convergent validity.

At the policy level a number of limitations of the OPUS are to be noted. It is recognized that the sample adopted

Acknowledgements

The project was funded by the Department of Health. Financial support from the University of Aberdeen and Scottish Executive Health Department is also acknowledged. Thanks to Andrew Healey, Martin Knapp, Til Wykes and Gavin Mooney for comments on this work. The views in the paper are those of the authors.

References (53)

  • D.G. Altman

    Practical Statistics for Medical Research

    (1991)
  • J.E. Araña et al.

    Emotions and decision rules in discrete choice experiments for valuing healthcare programmes for the elderly

  • A. Baddeley

    Working memory

    Philosophical Transactions of the Royal Society of Medicine London Series

    (1983)
  • M. Bradley

    Users Manual for the Speed Version 2.1 Stated Performance Experiment

    (1991)
  • P. Dolan

    Modelling valuations for Euroqol health states

    Medical care

    (1997)
  • C. Donaldson et al.

    Willingness to pay for health care

  • C. Donaldson et al.

    Should QALYs be programme specific?

    Journal of Health Economics

    (1988)
  • C. Donaldson et al.

    Assessing community values in health care: Is the ‘willingness to pay’ method feasible?

    Health Care Analysis

    (1997)
  • C. Donaldson et al.

    Willingness to pay: a method for measuring preferences for maternity care?

    Birth

    (1998)
  • Z. Hakim et al.

    Modelling the EuroQol data: a comparison of discrete choice conjoint and conditional preference modelling

    Health Economics

    (1999)
  • W. Hanemann

    Welfare evaluations in contingent valuation experiments with discrete responses: reply

    American Journal of Agricultural Economics

    (1984)
  • D.A. Hensher et al.

    Applied Choice Analysis: A Primer

    (2005)
  • R. Johnson et al.

    Willingness to pay for improved respiratory and cardiovascular health: a multiple-format, stated preference approach

    Health Economics

    (2000)
  • R. Johnson

    Willingness to pay, willingness to wait and super QALYs. RTI Health Solutions Monograph

    RTI Health Solutions

    (2005)
  • Kuhfeld, W., 1997. Efficient experimental designs using computerized searches, research paper series. Last visited June...
  • W. Kuhfeld

    Marketing Research Methods in the SAS System, Version 8 Edition

    (2000)
  • Cited by (142)

    • Test-Retest Reliability of EQ-5D-Y-3L Best-Worst Scaling Choices of Adolescents and Adults

      2023, Value in Health
      Citation Excerpt :

      Moderate test-retest reliability was found for adolescents in best choices, and fair reliability was found for adolescents in worst choices, with kappa ranging from 0.39 to 0.46. Compared with evidence reported previously by DCE (see Appendix Table 4.1 in Supplemental Materials found at https://dx.doi.org/10.1016/j.jval.2022.07.007 for detailed kappa results for DCE reported in previous studies in healthcare area), the kappa we reported suggests that BWS has comparable or slightly less reliability in adults.19,26-32 For example, Xie et al26 reported kappa of 0.528 for DCE with duration in valuing SF-6Dv2 health states.

    • Quality of care experience in aged care: An Australia-Wide discrete choice experiment to elicit preference weights

      2021, Social Science and Medicine
      Citation Excerpt :

      It has previously been applied in several studies assessing quality in the health care sector (de Bekker-Grob et al., 2012; Clark et al., 2014). It has also been successfully extended into estimating the value of social care for older people in the UK (Ryan et al., 2006; Burge et al., 2010), as well as aged care in Australia with a focus on older people and family carer preferences for a Consumer Directed Care model of community aged care service delivery (McCaffrey et al., 2015; Kaambwa et al., 2015), resident and family carer's preferences for the quality of care provided in nursing homes (Milte et al., 2018a) and resident and family carer preferences for food quality in nursing homes (Milte et al., 2018b). This study aimed to provide a better understanding of the Australian general public's perceptions of the key attributes that define the quality of care experience in aged care.

    View all citing articles on Scopus
    1

    Tel.: +44 1227 823644; fax: +44 1227 827038.

    2

    Tel.: +44 1224 551911; fax: +44 1224 550926.

    View full text