Comparison between exploratory factor-analytic and SEM-based approaches to constructing SF-36 summary scores

Abstract

To compare the two higher order factor structures of the Short-Form 36 (SF-36) Health Survey, using exploratory factor analytic methods and structural equation modeling (SEM). Two population data sets were used. A stratified representative sample (n = 1,005) of the Greek general population was approached for interview. This survey containing the SF-36 was used to obtain component score coefficients from principal components analysis and orthogonal rotation. These coefficients were then used in the second data set (n = 1,007) of the Greek adult general population to compute scores for the physical component summary and the mental component summary of the SF-36. The second data set was also used to obtain factor scores for physical and mental health measures, applying SEM. Exploratory factor analysis supported the existence of two principal components that are the basis for summary physical and mental health measures. SEM showed that models assuming that physical and mental health are correlated provided a better fit to the data than models assuming independence between physical and mental health. However, all eight dimensions of SF-36 should be included in the construction of summary scores. These results confirm the multidimensional structure of the SF-36, the correlational equivalence between standard summary measures and SEM-based second-order factor scores, and underscore the feasibility of multinational comparisons of health status using this instrument.