Objective: To develop and evaluate scaling and scoring assumptions, and the reliability and the validity of a Chinese version of the SF-36 scales.
Methods: A multi-stage mixed sampling procedure was used to select a representative sample of the general population. The sample size was 1 000 households. All family members of a selected household, aged 18 and older, completed a survey by self-administration. Formal psychometric methods for testing assumptions underlying item scoring and scale construction were used according to the standard procedure of the IQOLA Project.
Results: Of the 1 985 collected questionnaires, 1 972 were qualified. Of them, 1 688 (85.6%) were respondents. 1 316 respondents answered all 36 items, while the remaining (372 respondents) answered with one or several missing responses. Among the non-respondents, 65.5% were illiteracy or quasi-illiteracy. The assumption of equal intervals was violated for the VT and MH scales. The recoded item values were used to calculate scale scores. The clustering and ordering of the item means were approximately the same as that of the source version and other two Chinese versions. The items in each scale had similar standard deviations except those in the PF, BP, SF scales. The correlations between an item and its hypothesized scale were identical for all except the SF and VT scales. The scaling success rates of convergent validity were 0% for the SF scale, 75% for the VT scale, and 100% for the other six scales. The scaling success rates of discriminat validity ranged from 87.5% to 100% for all scales except for the SF scale. The Cronbach' alpha coefficients of internal consistency reliability ranged from 0.72 to 0.88, which were satisfactory for group comparison except 0.39 for the SF scale and 0.66 for the VT scale. The two-weeks test-retest reliability coefficient ranged from 0.66 to 0.94. Factor analysis identified two principal components: a "physical" factor and a "mental" factor. Taken together, these two factors could be used to explain 56.3% of the total variance.
Conclusion: The Chinese version of the SF-36 Health Survey Scale has achieved conceptual equivalence and satisfied the psychometric scaling assumptions well enough to warrant wide use in China. Known-groups validity will give more meaningful evidences of the validity of the Chinese SF-36 scales.