Psychometric Performance Comparison of the Adapted versus Original Versions of the EQ-5D-Y-3L and -Y-5L in Proxy Respondents for 2- to 4-Year-Olds.

Abstract

Few preference-weighted instruments are available to measure health-related quality of life in young children (2-4years of age). The EQ-5D-Y-3L and EQ-5D-Y-5L were recently modified for this purpose. The aim of this study was to test the psychometric properties of these adapted versions for use with parent proxies of children aged 2-4 years and to compare their performance with the original versions. It was hypothesised that the adapted instrument wording would result in improved psychometric performance. Survey data of children aged 2-4 years were obtained from the Australian Paediatric Multi-Instrument Comparison study. Distributional and psychometric properties tested included feasibility, convergence, distribution of level scores, ceiling effects, known-group validity (Cohen's D effect sizes for prespecified groups defined by the presence/absence of special healthcare needs [SHCNs]), test-retest reliability (intraclass correlation coefficients [ICCs]), and responsiveness (standardised response mean [SRM] effect sizes for changes in health). Level sum scores were used to provide summary outcomes. Supplementary analysis using utility scores (from the Swedish EQ-5D-Y-3L value set) were conducted for the adapted and original EQ-5D-Y-3L, and no value sets were available for the EQ-5D-Y-5L. A total of 842 parents of children aged 2-4 years completed the survey. All instruments were easy to complete. There was strong convergence between the adapted and original EQ-5D-Y-3L and EQ-5D-Y-5L. The adapted EQ-5D-Y-3L and adapted EQ-5D-Y-5L showed more responses in the severe levels of the five EQ-5D-Y dimensions, particularly in the usual activity and mobility dimensions (EQ-5D-Y-5L: mobility level 1: adapted n=478 [83%], original n=253 [94%]; mobility level 4/5: adapted n=17 [2.9%], original n=4 [1.5%)]). The difference in the distribution of responses was more evident in children with SHCNs. Assessment of known-group validity showed a greater effect size for the adapted EQ-5D-Y-3L and adapted EQ-5D-Y-5L compared with the original instruments (EQ-5D-Y-5L: adapted Cohen's D = 1.01, original Cohen's D = 0.83) between children with and without SHCNs. The adapted EQ-5D-Y-3L and adapted EQ-5D-Y-5L showed improved reliability at 4-week follow-up, with improved ICCs (EQ-5D-Y-5L: adapted ICC = 0.83, original ICC = 0.44). The responsiveness of all instruments moved in the hypothesised direction for better or worse health at follow-up. Probability of superiority analysis showed little/no differences between the adapted and original EQ-5D-Y-3L and EQ-5D-Y-5L. Supplementary psychometric analysis of the adapted and original EQ-5D-Y-3L using utilities showed comparable findings with analyses using level sum scores. The findings suggest improved psychometric performance of the adapted version of the EQ-5D-Y-3L and EQ-5D-Y-5L in children aged 2-4 years compared with the original versions.