Objectives: The expansion in descriptive system from the EQ-5D-3L (3L) to the EQ-5D-5L (5L) was largely motivated by an interest in greater discriminative ability as a measure of health in ill populations. However, the adequacy of the 5L descriptive system as a general population health measure has received limited attention. The primary aim of this study was to examine the properties of the 5L as a measure of population health in comparison to other variations in number of response levels.

Methods: Data were analyzed from the 2017 5L valuation study of a general population of US adults. Respondents self-reported their health using the 5L, 3L, and a 101-level visual analogue rating scale (VAS) for each dimension, with 0 representing extreme problems and 100 no problems. For insight on a more granular scale, response levels of the 5L were calibrated with mean VAS ratings across each dimension. The descriptive richness of variations on the number of levels was characterized by Shannon’s indices (H’; a measure of informativity and J’; a measure of descriptive efficiency controlling for number of levels).

Results: The dataset included a total of 3151 participants. Averaged across all 5L dimensions, mean (SD) VAS ratings by levels 1/2/3/4/5 were 95.5 (11.1), 79.4 (19.0), 63.4 (23.4), 47.6 (27.6), 48.1 (33.6), respectively. Only anxiety/depression demonstrated a substantial difference in mean VAS ratings between levels 4 and 5 (20.6, p<0.0001). Across dimensions, descriptive efficiency (J’) tended to increase when combining levels 4 and 5 into a 4-level system, with only marginal reductions in informativity (H’).

Conclusions: Results indicate that US general population respondents poorly differentiated between levels 4 and 5 for most dimensions when anchoring on an independent 101-level scale. In consideration of the balance between descriptive richness and efficiency, further research into applications for a 4-level EQ-5D measure may be warranted.