Are We Agreed? Self- Versus Proxy-Reporting of Paediatric Health-Related Quality of Life (HRQoL) Using Generic Preference-Based Measures: A Systematic Review and Meta-Analysis.

Objective: The aim of this study was to examine the level of agreement between self- and proxy-reporting of health-related quality of life (HRQoL) in children (under 18 years of age) using generic preference-based measures.

Methods: A systematic review of primary studies that reported agreement statistics for self and proxy assessments of overall and/or dimension-level paediatric HRQoL using generic preference-based measures was conducted. Where available, data on intraclass correlation coefficients (ICCs) were extracted to summarise overall agreement levels, and Cohen's kappa was used to describe agreement across domains. A meta-analysis was also performed to synthesise studies and estimate the level of agreement between self- and proxy-reported paediatric overall and domain-level HRQoL.

Results: Of the 30 studies included, 25 reported inter-rater agreement for overall utilities, while 17 reported domain-specific agreement. Seven generic preference-based measures were identified as having been applied: Health Utilities Index (HUI) Mark 2 and 3, EQ-5D measures, Child Health Utility 9 Dimensions (CHU9D), and the Quality of Well-Being (QWB) scale. A total of 45 dyad samples were included, with a total pooled sample of 3084 children and 3300 proxies. Most of the identified studies reported a poor inter-rater agreement for the overall HRQoL using ICCs. In contrast to more observable HRQoL domains relating to physical health and functioning, the inter-rater agreement was low for psychosocial-related domains, e.g., 'emotion' and 'cognition' attributes of both HUI2 and HUI3, and 'feeling worried, sad, or unhappy' and 'having pain or discomfort' domains of the EQ-5D. Parents demonstrated a higher level of agreement with children relative to health professionals. Child self- and proxy-reports of HRQoL showed lower agreement in cancer-related studies than in non-cancer-related studies. The overall ICC from the meta-analysis was estimated to be 0.49 (95% confidence interval 0.34-0.61) with poor inter-rater agreement.

Conclusion: This study provides evidence from a systematic review of studies reporting dyad assessments to demonstrate the discrepancies in inter-rater agreement between child and proxy reporting of overall and domain-level paediatric HRQoL using generic preference-based measures. Further research to drive the inclusion of children in self-reporting their own HRQoL wherever possible and limiting the reliance on proxy reporting of children's HRQoL is warranted.