Agreement and measurement properties of the interviewer-administered, self-completed and proxy-reported versions of the Tigrinya EQ-5D-Y-5L in Ethiopia

OBJECTIVE: To assess measurement properties of self-complete (SC), interviewer-administered (IA), and proxy-report versions of EQ-5D-Y-5L and to assess the degree of agreement in response between these versions. METHODS: SC, IA, and proxy-report were administered to children and adolescents with and without health conditions on two occasions, separated by either a 10-day interval (to assess test-retest reliability) or a one-month interval (to assess responsiveness). The SC and IA versions were administered in randomized order. Agreement and test-retest reliability assessed using Gwet's agreement coefficient (Gwet's AC2) and intraclass correlation coefficient (ICC) and, interpreted based on predefined thresholds. Feasibility, known-group validity, and responsiveness were assessed using missing values, Kruskal-Wallis test, and standardized effect size (SES), respectively. RESULTS: A total of 644 child/parent dyads participated. Missing values were higher for SC, particularly among younger children aged 8-12 years (4.65%-11.63%). Agreement between EQ-5D-Y-5L versions ranged from moderate to almost perfect, with excellent test-retest reliability across all dimensions and EQ-VAS. The EQ-5D-Y-5L level sum score (LSS) significantly differentiated between known disease groups across all versions (SC: χ² = 285; IA: χ² = 273; Proxy-report: χ² = 232; all p < 0.01). Both SC and IA versions were responsive to change in health with a moderate SES (SC: 0.74, - 0.61; IA: 0.36, - 0.66) in "worsened" and "improved" groups, respectively. CONCLUSION: The substantial agreement and similar measurement properties between SC and IA versions suggest that IA could be used as an alternative approach for data collection when self-complete is not an option. Although the proxy-report version had acceptable measurement properties, they were inferior to the SC and IA versions.