The estimation of a preference-based measure of health from the SF-36
Introduction
Measures of health related quality of life (HRQoL) have become widely used by clinical researchers and can provide useful descriptive information on the effectiveness of health care interventions covering such disparate range of outcomes for HRQoL. However, these measures have not been designed for use in economic evaluation. The main shortcoming of using such instruments in economic evaluation is that they do not explicitly incorporate preferences into their scoring algorithms.
This paper reports on a study to derive a preference-based measure of health from the SF-36, which is one of the most widely used generic measures of HRQoL in clinical trials. It has the potential to considerably extend the scope for undertaking economic evaluation in health care using existing and future SF-36 data sets. The paper also seeks to address the methodological issues this research task raises.
Section 2 of this paper briefly describes the SF-36 and some of the problems of using it in its current form in economic evaluation. This is followed by a section describing the methods of the study, including: the rationale for the choice of approach, the changes made to the SF-36, the valuation survey using a version of standard gamble (SG) and the issues around modelling the data. The valuation survey is reported in 4 The valuation survey, 5 The data and the modelling reported in Section 6. These types of stated preference data are complex to model due to their hierarchical nature and skewed distribution, and Section 6 outlines alternative specifications for dealing with these problems. The final section considers how the results from this work can be used.
Section snippets
The short form-36 (SF-36) health survey
The SF-36 health survey is a standardised questionnaire used to assess patient health across eight dimensions (Ware et al., 1993). It consists of items or questions which present respondents with choices about their perception of their health. The physical functioning dimension, for example, has 10 items to which the patient can make one of three responses: ‘limited a lot’, ‘limited a little’ or ‘not limited at all’. These responses are coded 1, 2 and 3, respectively, and the 10 coded responses
Methods
There are three components to this study. Firstly, the SF-36 has been reduced in size and complexity in order that respondents can process the information and hence give reliable valuations of health states. Secondly, a preference based valuation survey has been undertaken. Thirdly, the results of the survey were used in a model to predict values for all states of health described by the reduced form version of the SF-36, via alternative econometric techniques.
Econometric methods to estimate a
The valuation survey
The basic design of the survey was that a sample of 249 health states defined by the SF-6D was valued by a representative sample of the general public (n=836). Each respondent was asked to rank, and then value, six of these states using a variant of the SG technique.
The data
Out of the 1445 addresses contacted for interview, 167 proved to be ineligible.1 Of the usable addresses there were 836 successfully conducted interviews (a 65% response rate). Respondents were found to be representative of the national population in terms of the distribution
Modelling
The overall aim is to construct a model for predicting health state valuations based on the SF-6D. The appropriate modelling strategy is not clear a priori, and the econometric analysis is necessarily of an exploratory nature (Busschbach et al., 1999). The data generated by the valuation survey described above, has a complex structure which creates a number of problems for econometric estimation. Firstly, the data are skewed and bimodal (see Fig. 1). Conventional power transformations are
Discussion and conclusion
The results of this study offer a method for analysing existing SF-36 data from trials and other sources of evidence where there is no other means of estimating the preference-based health values for generating QALYs. It also provides an alternative to existing preference-based measures of health for use in cost utility analysis. Two of the leading preference-based measures are the EQ-5D (Brooks, 1996) and the health utility index (Torrance et al., 1995). Whether or not the SF-6D offers an
Acknowledgements
We would like to thank GlaxoWelcome for supporting this study and Roger Thomas and Patrick Sturgis at SCPR for conducting the valuation survey. The usual disclaimer applies.
References (25)
EuroQol: the current state of play
Health Policy
(1996)- et al.
Estimating parametric relationships between health state description and health valuation with an application to the EuroQol EQ-5D
Journal of Health Economics
(1999) - Abdalla, M., Russell, I., 1995. Tariffs for the Euroqol health states based on modelling individual VAS and TTO data of...
- et al.
Deriving a preference based single index measure from the SF-36
Journal of Clinical Epidemiology
(1998) - Brazier, J.E., Deverill, M., Harper, R., Booth, A., 1999a. A review of the use of health status measures in economic...
- Brazier, J.E., Roberts, J., Deverill, M., 1999b. The estimation of a utility based algorithm from the SF-36 health...
- Breen, B., 1996. Regression Models: Censored, Sample Selected or Truncated Data. Sage,...
Modelling valuation for Euroqol health states
Medical Care
(1997)- Dolan, P., Gudex, C., Kind, P., Williams, A., 1995. A social tariff for Euroqol: results from a UK general population...
- Furlong, W., Feeny, D., Torrance, G.W., Barr, R., Horsman, J., 1990. Guide to design and development of health state...