Implications for research and clinical use from a Rasch analysis of the HOOS-12 and KOOS-12 instruments

Objective: To evaluate the structural validity of the 12-item Hip disability and Osteoarthritis Outcome Score (HOOS-12) and 12-item Knee injury and Osteoarthritis Outcome Score (KOOS-12) using Rasch analysis and consider psychometric implications for research and clinical use. Method: Individual-level HOOS-12 and KOOS-12 data from the Australian Orthopaedic Association National Joint Replacement Registry, collected before and after primary total hip and knee replacement, were used for this analysis. Using the Rasch analytic approach, overall model fit and item fit were examined, together with potential reasons for misfit including response threshold ordering, differential item functioning, internal consistency, unidimensionality and item targeting. Results: Overall misfit to the Rasch model was evident for both instruments. A degree of item misfit was also observed, although most items demonstrated logical sequencing of response options. Only two items (hip/knee pain frequency and awareness of hip/knee problems) displayed disordered response thresholds. The pain, function, and quality of life domains of the HOOS-12 and KOOS-12 demonstrated excellent internal consistency reliability (person separation index: 0.80–0.93) and unidimensionality. A mismatch between item difficulty and person ability scores at the highest end of the HOOS-12 and KOOS-12 scales contributed to post-operative ceiling effects (mean logit for HOOS-12: 3.57; KOOS-12: 2.58; ≈0 indicates well-targeted scale). Conclusion: We found evidence to support the structural validity of the three HOOS-12 and KOOS-12 domains for evaluating joint replacement outcomes. However, there may be missing content in both instruments particularly for high-functioning patients. Minor refinement of some response options may be warranted to improve item performance.