VALIDATION DATA

The internal consistency of the KOOS-PS in the original article was 0.90 and it represents a unidimensional construct by virtue of the fit of the data to the Rasch model

Since the KOOS-PS is not unidimensional, the internal consistency (pooled outcome of the Cronbach's alpha (0.85 in 3212 patients) is difficult to interpret and could not be used and the overall rating was scored as indeterminate.

For patients undergoing a total knee replacement, the internal consistency was 0.89.

 Pooled ICC of the KOOS-PS was 0.81 (95% CI 0.67–0.87) in 291 patients.

Seven items remained, each showing good fit to the model. The overall model χ2 indicated good fit to the Rasch model, with a χ2 of 73.34 and P=0.1751.

Principal component analysis of the residuals did not reveal any detectable patterns or systematic information, suggesting that the retained items formed a unidimensional structure. Rotated factor loadings showed that no more than one item loaded on each of the components, indicating that none was highly correlated with the others. Each item had a factor loading>0.90 on their respective component and loadings<0.22 on all other components. Further confirming the unidimensional structure, the t-test comparing person scores, estimated using the two sets of items distinguished by positive and negative factor loadings from the residual analyses, indicated that only 3.6% of observations lay outside the 95% CI.