The application of anchoring vignettes to the EQ-5D-5L: a possible solution to reporting heterogeneity in PROMs

Abstract: Self-reported health measures which use categorical responses, including patient reported outcome measures (PROMs), are integral components of healthcare decision-making. However, as these measures are by their nature subjective, response categories can often mean different things to different people. Systematic differences in the way that people use and interpret response categories, so called reporting heterogeneity or differential item functioning (DIF), can introduce bias when using self-reports to compare health across heterogeneous patient or population groups. This presentation will focus on a programme of research which has sought to test the feasibility of using anchoring vignettes to identify and adjust for DIF in EQ-5D-5L responses. The design of vignettes will be discussed, particularly in relation to achieving the identifying assumptions of response consistency and vignette equivalence. The results of an exploratory analysis, using 4,300 responses to online surveys in Australia, show that DIF can be identified (at least in certain age groups), and that adjustments for DIF have significant effects on the inference of EQ-5D differences. The presentation will also provide results with respect to using vignettes to correct for DIF in an external dataset (that is one in which vignette responses were not collected). Anchoring vignettes have frequently been described as “the most promising” approach for detecting DIF, this programme of research has confirmed the promise holds also for the most commonly used PROM, the EQ-5D.