36PBP GLOBAL VALUES FOR EQ-5D HEALTH STATES : AN INNOVATIVE APPLICATION OF META-ANALYSIS

Sunday, October 18, 2009
Grand Ballroom, Salons 1 & 2 (Renaissance Hollywood Hotel)
Victor Zarate, MD, MSc, Paul Kind and Ling-Hsiang Chuang, MSc, University of York, York, United Kingdom

Purpose: During the last decades researchers have carried out valuation exercises which have allowed the estimation of EQ-5D social value sets in different countries based on a valuation protocol originally developed in the UK. The preferred valuation methodology for collecting these social preferences has been Time Trade-Off (TTO), particularly when the estimation of health utilities is required in economic evaluations. The objective of this study is to develop a global EQ-5D value set through a meta-analysis of previously reported valuation studies.

Methods: Based on a literature review, all EQ-5D valuation studies performed since 1993 were selected for a preliminary analysis. Only those which reported number of observations, mean and standard deviation for observed TTO values were included for meta-analysis. Random effect summary estimates were calculated through inverse-variance weighting using the UK MVH study as the reference/control group. Mean differences between several countries and the UK were standardised by the Glass method (SMD) and graphically described by forest plots, one for each of the 42 EQ-5D health states commonly used in valuation studies. Consistency of the results was assessed by the I-squared term.

Results: Out of 11 national EQ-5D valuation studies, 7 countries were selected for further analysis: Chile, USA, Poland, Japan, Korea, UK and Germany. In general, all summary estimates tend to be higher than TTO values reported in the UK. When 95%CI of the SMD were taken into account, this relation was observed in 29/42 health states. The mean absolute difference between UK values and summary estimates was 0.15, ranging between 0.004 and 0.399 for EQ-5D health states 11211 and 11133 respectively. Consistency between studies, as measured by I-square, ranged between 82.7 and 99%. Using summary estimates, a value set for 42 health states was computed. These data were used in an OLS estimation model to determine values for all 243 EQ-5D health states.  Explained variance was predictably high (r2 > 0.9).

Conclusions: There is a high level of heterogeneity between different valuation studies conducted around the world.  Summary TTO estimates tend to be higher than the original values reported in the UK. Given a lack of EQ-5D value sets in many countries, these findings provide a global value set for EQ-5D that can be used for international economic or public health applications.

Candidate for the Lee B. Lusted Student Prize Competition