Cross-cultural adaptation of a psychometric instrument: two methods compared

Cross-cultural adaptations of questionnaires are needed in multilingual research, but little is known about the effectiveness of specific translation methods. We compared properties of two French-language adaptations of the SF36 health survey: (a) a rapid translation developed over 3 months in Geneva in 1992 (Geneva version), based on three initial translations, one synthesis, and two pretests, and (b) a comprehensive adaptation developed by the International Quality of Life Assessment Project between 1991 and 1994 (IQOLA version), which involved back-translations, focus groups, development of equidistant response options, item difficulty and quality ratings, and multiple pretests. Wordings of 34 of 36 items differed. These two instruments were administered 1 year apart to the same sample of 946 young adults. Ceiling effects were somewhat lower for the IQOLA than for the Geneva version (means 30.4% and 35.5%), and missing scores slightly less frequent (IQOLA: mean 0.5%; Geneva: 1.2%). Floor effects (means 2.7% and 2.4%), proportions of consistent respondents (93.4% and 94.0%), and internal consistency coefficients (IQOLA: 0.78-0.89, Geneva: 0.80-0.92) were similar. Factor analysis supported the existence of two main aspects of health (physical and mental) for both versions. A majority of known-group comparisons were compatible with theory, for both versions. In conclusion, the two French-language versions of the SF36 had similar psychometric properties, despite extensive differences in the development process. This suggests that a moderately resource-intensive translation may produce adequate results. More empirical research is needed to understand what translation methods yield the best results.