Differential item functioning in the EQ-5D: An exploratory analysis using anchoring vignettes

Inter-group comparisons using the EQ-5D, or any self-reported measure of health, rely on the measure being an accurate reflection of the true health of the groups or individuals concerned. However, responses to questions on subjective scales, such as those used in the EQ-5D, will be inaccurate if groups of individuals systematically differ in their use of the response categories, a phenomenon known as differential item functioning (DIF). This paper reports on an exploratory analysis involving the use of anchoring vignettes to identify differential item functioning (DIF) in the EQ-5D-5L. We demonstrate that using vignettes to appropriately identify DIF in EQ-5D reporting is possible, at least in certain age groups. We find that the EQ-5D is indeed subject to DIF, and that failure to account for DIF can lead to conclusions that are misleading when using the instrument to compare health or quality of life across heterogeneous groups. For instance, when adjusting for DIF in a sample aged 55-65 years, we found that differences between the highest and lowest education groups doubled in value after adjusting for DIF, and increased from quantities that would not have had relevance in a clinical settings to ones that would (based on a suggested minimally important difference). Thus, our research provides evidence that the EQ-5D should be used with caution when comparing health or quality of life across heterogeneous groups. We also provide several important insights in terms of the identifying assumptions of response consistency and vignette equivalence. Acknowledgements: This research was funded by an Australian Research Council Discovery Project Grant (DP110101426), a BankWest Curtin Economics Centre (BCEC) grant, and a Monash Faculty of Business and Economics grant.