EQ-5D-3L index scores in patient and general populations typically have a non-normal distribution, divided into 2 distinct groups. It is important to understand to what extent this is determined by the way that the EQ-5D-3L index is constructed rather than by the true distribution of ill health.
This paper examines the determinants of the “2 groups” distribution pattern and the extent to which this pattern is attributable either to the EQ-5D-3L classification system used to create health state profiles or to the weights applied to profiles.
Data from the English NHS PROMs program (hip and knee replacements and varicose vein and hernia repairs) and from a study of 2 chronic conditions (asthma and angina) were used to compare the distributions of EQ-5D-3L index scores with distributions from which weights have been stripped; profile data decomposed into their constituent dimensions and levels; a condition-specific index; and using weights from different countries, based on both time tradeoff and visual analogue scale.
The EQ-5D-3L classification system generates differences between patients with the same condition in respect of dimensions that are mainly observed at level 2 or 3. The weights commonly used to calculate the index exacerbate this grouping by placing a larger weight on level 3 observations, generating a noticeable gap in index scores between the groups.
Analyzing EQ-5D profile data enables a better understanding of the resulting distribution of EQ-5D scores. The distinctive shape observed for these distributions is the result of both the classification system and the weights applied to it.