Statistical Analysis of EQ-5D Profiles: Does the Use of Value Sets Bias Inference?

Abstract

Health state profile data, such as those provided by the EQ-5D, are widely collected in clinical trials, population surveys, and a growing range of other important health sector applications. However, these profile data are difficult to summarize to give an overall view of the health of a given population that can be analyzed for differences between groups or within groups over time. A common way of short cutting this problem is to transform profiles into a single number, or index, using sets of weights, often elicited from the general public in the form of values. Are there any problems with this procedure? In this article, the authors demonstrate the underlying effects of the use of value sets as a means of weighting profile data. They show that any set of weights introduces an exogenous source of variance to health profile data. These can distort findings about the significance of changes in health between groups or over time. No set of weights is neutral in its effect. If a summary of patient-reported outcomes is required, it may be better to use an instrument that yields this directly (such as the EQ VAS) along with the descriptive instrument. If this is not possible, researchers should have a clear rationale for their choice of weights and be aware that those weights may exert a nontrivial effect on their analysis. This article focuses on the EQ-5D, but the arguments and their implications for statistical analysis are relevant to all health state descriptive systems.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 98 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.