Combining four patient-reported outcome measurement instruments to develop a common metric computerized adaptive test to measure outcome after total knee arthroplasty.

Abstract

To map the Oxford Knee Score (OKS), OKS Activity & Participation Questionnaire, Forgotten Joint Score (FJS), and High Activity Arthroplasty Score to a common scale using principles of modern test theory. Using the common scale, we then aimed to build a computerized adaptive test (CAT) to reduce item burden. Participants undergoing total knee arthroplasty provided preoperative and postoperative (six and 12 months) paired responses for four patient-reported outcome measure (PROM) instruments, with the OKS being the common linking instrument. After meeting assumptions (unidimensionality, monotonicity, local independence, and measurement invariance), a graded response model was fitted to the data. A CAT algorithm was developed. There were 6,301 paired responses included. Confirmatory factor analysis demonstrated that the four PROM instruments measured a similar or very similar latent construct, which we describe as 'knee health'. The combination of all four instruments resulted in higher test-level information than any individual instrument. The CAT reduced 36 items to a median of eight items, while maintaining high measurement precision (standard error of measurement 0.29). The common metric provides a more precise measurement of knee health than any of the individual instruments, and avoids floor and ceiling effects inherent in individual scores. The CAT algorithm reduced the item burden to a median of eight items, which is lower than the FJS or OKS, when used as individual instruments. It will also allow for standardization of outcome reporting and pooling of results across databases that use any of the four instruments to facilitate individual and meta-analyses across different scores.