• Sessions
• Presentations
• Posters
• Rooms

Presentation

Session: Posters
Room: TBA
Time: Fri 13:00-14:30

Presenter: Leidl Reiner (German Research Center for Environmental Health . Institute for Health Economics and Health Care Management)

Abstract

Aim: In cases where the valuation of health states is not available on the basis of the target population or the valuation approach that the decision makers require, index models can be used instead. Index models are typically based on valuations of given health states and often claimed to render decision utilities. The approach proposed here is to base the index on valuations of experienced health states, and to compare the performance of these two basic approaches.

Data and Methods: For health state description and for their valuation by a rating scale, the EQ-5D is used. Data are retrieved from representative German population surveys. Using a 2006 survey, the index is estimated by (i) a generalized linear model (GLIM) with binomial error distribution and parameter constraints. Comparisons are made with two existing German indices that were derived from valuations of given health states in other population samples: (ii) a time-trade off valuation, and (iii) a rating scale valuation. For comparison of performance of these indices, the actual ratings of health states in a 2007 survey serve as the gold standard. Performance is measured by the correlation of actual and predicted values as well as by measures of distance between the two. Performance is tested both in the whole 2007 population sample and in a sub-sample of respondents who reported to have diabetes.

Results: The GLIM approach avoids the emergence of inconsistent parameters. Basing valuations on experienced health states restricts the number of very severe health states that enter the index estimation, but increases the number of less severe states. In terms of predictive accuracy in the total population sample, the new approach performs equal to, and partly better than the two existing approaches. Further improvement in accuracy is feasible by integrating age and sex adjustments in the estimation. Predictive accuracy in the sub-sample of respondents with diabetes was clearly reduced for the GLIM approach, but much less so than the reduction in predictive accuracy for the two existing approaches.

Discussion: Several shortcomings of the traditional econometric approach can be coped with by the proposed GLIM approach. Main limitations of this approach concern a restricted range of health states covered, and performance so far only having been tested in a general population sample. First results with respect to respondents with diabetes indicate that prediction in clinical sub-groups may be less accurate than for the general population. Yet, the GLIM approach proved better predictive accuracy there than the existing approaches.

Conclusion: The approach to estimate a population-relevant index for the EQ-5D from experience-based valuations renders plausible first results. This introduces a new option for decision makers who prefer to rely on valuations of experienced health states rather than valuations of hypothetical ones. Further testing of the GLIM approach is needed on whether or not appropriate performance can be claimed for in clinical populations.

Key Terms

Valuing health, GLIM, performance measurement

Authors:

Leidl Reiner (Helmholtz Zentrum München. Institute for Health Economics and Health Care Management) and Reitmeir Peter (Helmholtz Zentrum München. Institute for Health Economics and Health Care Management)

Search with Google