An Investigation of Inter-Rater and Intra-Proxy Agreement in Measuring Quality of Life of Children in the Community Using the EQ-5D-Y-3L

Background: Self-reporting of health-related quality of life (HRQoL) in children is not always feasible. To date, proxy perspectives (Proxy versions 1 and 2) using the EQ-5D-Y-3L have not been explored for its impact on agreement with child self-report. Proxy version 1 requires the proxy to consider their own view of the child’s HRQoL (proxy-proxy), while with Proxy version 2, the proxy is asked to respond as they believe their child would self-report their HRQoL (proxy-child). This study compared the inter-rater and intra-proxy agreement (overall and dimension level) using the EQ-5D-Y-3L self, proxy-proxy, and proxy-child reports. 

Methods: A community-based sample of child (aged 6–12 years) and parent dyads were invited to participate in a semi-structured interview. The child self-completed the EQ-5D-Y-3L independently of the parent who completed the EQ-5D-Y-3L from proxy-proxy and proxy-child perspectives. Agreement was determined using Concordance Correlation Coefficients (CCCs) for the overall (preference-weighted) HRQoL, while agreement at the dimension level was evaluated using Gwet’s agreement coefficient (AC₁). To assess the differences between the self and the two proxy reports, the Wilcoxon matched-pair signed-rank test was used. 

Results: This study involved 85 child-parent dyads. The agreement between self and proxy overall HRQoL was low (fair) with both proxy-proxy (CCC = 0.28) and proxy-child (CCC = 0.26) reports. The largest discrepancy in the child-proxy agreement at dimension level with both the proxy versions was observed for ‘feeling worried, sad or unhappy’. Within this dimension, the proxy-child perspective resulted in a stronger agreement (AC₁ = 0.7, good) with child self-report compared with the traditional proxy-proxy perspective (AC₁ = 0.58, moderate). Although the preference-weighted HRQoL was consistent across both the proxy perspectives, a significant difference was observed in the EQ VAS scores (p = 0.02). 

Conclusions: This study demonstrates that choice of proxy perspective may have an impact on the problems reported on HRQoL dimensions and EQ VAS scores. However, in this community-based sample of generally healthy children, no significant difference was observed in the inter-rater agreement for child-self and proxy preference-weighted EQ-5D-Y-3L values based on proxy perspectives. While this suggests that preference-weighted data are not sensitive to the choice of perspective, these findings may differ for different HRQoL instruments and for alternative value sets with different properties.