Measurement invariance of the SF12v2 between its English and Chinese versions in Chinese adults

Abstract

Aims: The SF-12v2 has been a popular quality of life assessment tool across different cultural groups. However, measurement invariance between its English and Chinese versions has not been robustly assessed. Therefore, we aimed to assess the measurement invariance between the English and HK Chinese versions of the SF-12. Methods: We conducted a university-based survey of individuals aged 18 years of above. Individuals who were bilingual in both English and Chinese were randomized to complete either the English or Chinese version of the SF-12v2. After recoding the item responses as stipulated in the scoring manual, the two-component scale model comprising the physical component and mental component scales was assessed in a confirmatory factor analysis (CFA). Comparative fit index (CFI), Tucker-Lewis Index (TLI). Root Mean Square Error of Approximation (RMSEA), and Standardized Root Mean Square Residual (SRMR) were used to assess model fit. Then, multiple-group CFA was used to assess configural invariance when the same scale model was imposed in both language groups, weak invariance when the factor loadings were constrained to be equal, and strong invariance when item intercepts were further constrained to be the same. Results: A total of 1,358 bilingual participants responded to all items of the SF-12v2. Of which, 590 completed the English version, and 795 completed the Chinese version. Their mean age of 22 years (SD: 6, Range 18–68), with 861 were female. The 2-component model of the SF-12v2 showed satisfactory fit in all participants (Χ2 = 444.9, df = 48, RMSEA = 0.077, CFI = 0.98, TLI = 0.97, SRMR = 0.054). Imposing the multiple group by language showed satisfactory fit (Χ2 = 505.7, df = 96, RMSEA = 0.079, CFI = 0.98, TLI = 0.97, SRMR = 0.056). Constrained equal factor loadings resulted in significant decrease in fit (Χ2 = 63, df = 10, p < 0.001) but only small change in CFI (0.008). Further constrain of equal item intercepts also resulted in significant decrease in fit (Χ2 = 243.5, df = 10, p < 0.001) with change of CFI by 0.037. Conclusions: The SF12-v2 reached weak measurement invariance between its English and Chinese versions. Differential item functioning may be explored before using the SF12v2 in studies involving both English and Chinese languages.