A cross-country comparison of the psychometric performance of SF-6Dv2 and EQ-5D-5L

To evaluate the comparative performance of SF-6Dv2 and EQ-5D-5L in the general population of Quebec (Canada), Tehran (Iran), and Japan.

Data on SF-6Dv2 and EQ-5D-5L were collected in the three countries. Descriptive differences in utility values between SF-6Dv2 and EQ-5D-5L were assessed using t-tests, as well as ceiling effects were evaluated based on the percentage of "no problem" levels reported. The known-group validity of both measures was assessed by comparing utility scores across health and demographic subgroups using t-tests or ANOVA and by calculating effect sizes across known groups. The area under the receiver operating characteristic curve (AUROC) analysis and F-statistic ratios were used to further validate the findings from the known-group validity analyses. Convergent validity for both instruments was assessed using Spearman's rank correlation coefficient. The agreement between instruments was evaluated using intraclass correlation coefficients (ICC) and Bland-Altman plots.

A total of 2,378 respondents for Quebec, 3,061 for Tehran, and 3,933 for Japan were included. Differences in utility values between SF-6Dv2 and EQ-5D-5L were statistically significant, with SF-6Dv2 generally yielding lower utility scores. Both instruments demonstrated strong known-group validity, effectively distinguishing between diseased and healthy groups as well as various demographic characteristics. However, EQ-5D-5L outperformed SF-6Dv2 for most demographic characteristics based on AUROC analysis and F-statistic ratios. In contrast, their performance in distinguishing between healthy and diseased groups did not favor a particular instrument. Convergent validity analyses indicated strong associations between SF-6Dv2 and EQ-5D-5L utility values in Quebec (0.760) and Tehran (0.737). The agreement between SF-6Dv2 and EQ-5D-5L utility values was moderate in Quebec (0.69) and strong in Tehran (0.837). Bland-Altman plots indicated that differences between the two instruments tended to increase as the average score decreased.

Both EQ-5D-5L and SF-6Dv2 demonstrated favorable psychometric performance in terms of known-group validity and convergent validity. These findings suggest that both instruments are valid tools for health utility measurement for use in general population.