Objectives: To evaluate the performance of the first non-English (Swedish) SF-36 version 2.0 (V2) regarding scaling assumptions, reliability and validity, with special emphasis on the effects of extending the response scales of the two role functioning scales, role physical (RP) and role emotional (RE), from a dichotomous to a five-step format.
Methods: Questionnaires were mailed to a non-stratified, random national sample of 3000 18-75 year old Swedish residents in 1998-1999. Methods traditionally used in evaluating the original US SF-36 version 1.0 (V1) and other international versions were applied.
Results: A total of 73% of the questionnaires were returned. Scaling assumptions were satisfactorily met and generally enhanced compared with V1, particularly regarding the role functioning scales. Floor/ceiling effects were reduced and reliability estimates increased for the role scales. The factor structure was replicated and the relative validity of the role scales as measures of their components increased. Overall, the tests of the criterion-based validity using known groups comparisons produced results supporting hypotheses: scales from the physical domain distinguished best between groups expected to differ in physical health; and scales comprising the mental domain distinguished best between groups expected to differ in mental health. Furthermore, hypothesized differences in mean scale scores as a function of age and social risk factors were confirmed.
Conclusions: The changes to the response formats of the role functioning scales have improved their precision, reliability and validity without jeopardizing the underlying structure of the original SF-36. These changes will likely further enhance the responsiveness of the SF-36.