OBJECTIVES: The growing variety of Health-Related Quality of Life (HRQoL) instruments and value sets calls for methods and tools to assess which alternative is most sensitive to change in a particular target population. We introduce novel methodology and an R package 'valueSetCompare', for graphical and statistical comparison of the sensitivity of different instruments and different value sets and demonstrate its usefulness by comparing the US EQ-5D-5L, SF-6D, and HUI3 value sets using data from the Multi Instrument Comparison (MIC) study and Visual Analogue Scale as a severity proxy.

METHODS: The graphical comparison relies on cross-sectional data incorporating utility values from two or more different instruments, and an independent variable correlated with severity. A large number of simulated datasets are generated using resampling where the likelihood of sampling is a function of the severity proxy score. Mean scores and confidence bands along the severity for each instrument are presented in a ribbon plot, where the slope represents sensitivity. Discriminative ability is also investigated using the ratio of F-statistics estimated from analysis of variance (ANOVA).

RESULTS: Graphical comparison indicates that EQ-5D-5L and HUI3 had greater sensitivity to severity changes than SF-6D. No statistically significant differences were observed using the F-statistic approach. Theoretical comparison revealed that EQ-5D-5L had the widest scale range (1.57) compared to HUI3 (1.35) and SF-6D (0.98). In EQ-5D-5L, 20% of health states were rated below 0, compared to none in SF-6D and 77.5% in HUI3.

CONCLUSIONS: The novel framework and the ‘valueSetCompare’ R package allow for intuitive graphical representation of the sensitivity of different instruments, as well as standardized statistical tests. Comparison of sensitivity could be used to determine which instrument is best suited to capture treatment effects in populations of interest.