The structural validity of the Danish version of the Oxford Knee Score is not substantiated using Rasch analysis and confirmatory factor analysis

AIMS: The Oxford Knee Score (OKS) is a 12-item patient-reported outcome measure (PROM), developed for patients who are candidates for total knee arthroplasty (TKA). A prerequisite for a PROM to be considered an adequate measurement instrument is a unidimensional structure as demonstrated by a modern test theory (MTT) model, meaning that each (sub)scale reflects one construct (e.g. pain). However, the structural validity of OKS has only been sparsely evaluated with MTT, and with ambiguous results. This study aimed to assess the structural validity of the Danish OKS. Since the OKS includes items addressing both pain and physical function, it was hypothesized that scores were more accurately reported as two separate subscales.

METHODS: OKS responses from a study of 1,059 patients treated with a TKA were obtained. Four random subsamples (each with 400 patients) - female and male, preoperative and three months postoperative - were assessed by confirmatory factor analysis (CFA) and Rasch analysis. CFA model fit was evaluated using the chi-squared statistic and indices of close fit. Rasch fit was evaluated with item fit statistics. Both a one-factor solution and two-factor solutions with scores based on two separate subscales were considered for each subsample.

RESULTS: OKS data did not fit the original unidimensional model of one total score. Reporting OKS data as the two subscales "pain" and "function" improved CFA fit, but model fit was still inadequate. Results were consistent across subsamples.

CONCLUSION: The structural validity of the Danish OKS is inadequate for evaluating patients awaiting TKA or surgically treated with TKA. OKS data should therefore be interpreted with caution. Randomized treatment studies showing no difference in OKS scores may be reanalyzed based on the two domains to reduce the risk of a potential type-2 error.