Dimensional structure of the SF-36 in neurological patients

Items were considered to load on (were related to) a factor if factor loadings exceeded 0.40 on that factor and were lower on the other factors. Because all items had ordinal scores and several subscale scores showed a skewed distribution, PCA was carried out using polychoric correlations instead of Pearson correlations [11]. Other authors also found proportions of explained variance below 60% and deviating factor contents for the subscales with the physical and mental summary scores in patients with MS (9), stroke (8), and ALS (10) and raised questions about the use of the SF-36 summary scales in these populations.