Differential item and test functioning methodology indicated that item response bias was not a substantial cause of country differences in mental well-being

Abstract

ObjectivesEstablishing the cross-cultural equivalence of the mental well-being construct, as measured with the Warwick-Edinburg Mental Well-being Scale (WEMWBS), by studying potential construct validity biases in two countries with previously reported score differences. Study Design and SettingWe compared the WEMWBS total scores and item responses in Scotland (N = 779) and Catalonia (N = 1,900) general population samples. To assess whether the questionnaire spuriously favored higher scores in Catalonia, we tested for differential item functioning (DIF) by applying ordinal logistic regression on Item Response Theory scores. DIF was tested with likelihood ratio tests and standard effect measures (McFadden Pseudo R2, >0.13; relative parameter change, >5%), and differential test functioning (DTF) was tested by plotting differences between full-test and purified (i.e., without DIF items) score estimates. ResultsCatalonia showed higher levels of mental well-being than Scotland (Cohen d = 0.84). Three of 14 WEMWBS items showed small amounts of DIF. DIF did not accrue to DTF, as shown by intraclass correlation coefficient (ICC, 0.999) and case-by-case differences (maximum, 0.12 SD) between total and purified scores. Population differences remained mainly constant across sociodemographics and health outcomes. ConclusionThe WEMWBS measures a distinct well-being construct that is stable across countries, implying that Scotland and Catalonia populations are effectively different in the distribution of mental well-being. This result adds to previous psychometric information and supports WEMWBS as a valid unbiased measures for individual and cross-cultural comparisons.