Objective: To assess the suitability for use within economic evaluation of a widely used sleep-related instrument (the Epworth Sleepiness Scale [ESS]) by examining its convergent and discriminant validity with two widely used generic preference-based instruments (Short-Form 36 [SF-36] and the Assessment of Quality of Life 4 dimensions [AQoL-4D]).
Methods: Data from a cross-section of 2,236 community-dwelling Australian men were analyzed. Convergent validity was investigated using Spearman's correlation, intraclass correlation, and modified Bland-Altman plots, while discriminant validity was examined using Kruskal Wallis tests.
Results: All instruments showed good discriminant validity. The ESS was weakly correlated to the Short Form 6 dimension, or SF-6D (derived from the SF-36) and AQoL-4D utilities (r = 0.20 and r = 0.19, respectively). Correlations between ESS and SF-36/AQoL-4D dimensions measuring the same construct were all in the hypothesized directions but also weak (range of absolute r = 0.00 to 0.18). The level of agreement between the ESS and AQoL-4D was the weakest, followed by that between the ESS and SF-6D. Moderate convergent validity was seen between the utilities.
Conclusions: The lack of convergent validity between the ESS and the preference-based instruments shows that sleep-related constructs are not captured by the latter. The ESS has, however, demonstrated good discriminant validity comparable to that of the AQoL-4D and the SF-36/SF-6D and would therefore be equally useful for measuring subgroup differences within economic evaluation. We therefore recommend using the ESS within cost-effectiveness analysis as a complement to preference-based instruments in order to capture sleep-specific constructs not measured by the latter.