Convergent and discriminant validity of the Barthel Index and the EQ-5D-3L when used on older people in a rehabilitation setting

This study compares the empirical performance of a commonly used functional-status measure, the Barthel Index (BI), to that of a widely used generic preference-based instrument, the EuroQoL-5-Dimensions 3 Level (EQ-5D-3L), in older people. Data from older people receiving rehabilitation services were used to test the validity of the BI and EQ-5D-3L. Convergent validity was investigated using Spearman’s correlation, exploratory factor analysis (EFA), scatter plots, Krippendorff’s alpha and modified Bland-Altman plots. Discriminant validity was examined using Kruskal Wallis tests, ceiling effects and EFA. A total of 1690 participants were included in the analysis. The BI total and EQ-5D-3L utility scores showed moderate correlation (r = 0.51; Krippendorff’s alpha = 0.52). Kendall’s Tau-B correlations between BI items and EQ-5D-3L dimensions measuring the same construct were weak to moderate (0.05 ≤ absolute r ≤ 0.54). In the EFA, some BI items cross-loaded onto the same factors as EQ-5D-3L dimensions, suggesting that the instruments were interrelated. The BI, however, focuses more on physical functioning, while the EQ-5D-3L measures broader wellbeing concepts. Both instruments showed good discriminant validity and would therefore be equally valuable for measuring subgroup differences. Researchers should consider using the BI in rehabilitation to capture more physical functioning-specific constructs not measured by the EQ-5D-3L.