Rasch analysis of the 22 knee injury and osteoarthritis outcome score-physical function items in Italian patients with knee osteoarthritis

Objectives: To replicate previous Rasch analysis (RA) that yielded the 7-item Physical Function Short Form of the Knee Injury and Osteoarthritis Outcome Score (KOOS-PS7) in order to verify the consistency of the item selection process, and to analyze its main psychometric properties in a sample of Italian patients with osteoarthritis (OA).

Design: Methodological research on cross-sectional data from a convenience sample.

Setting: University hospital.

Participants: Outpatients with symptomatic OA of the knee (N=200; women, 73%; mean age, 69.4y).

Interventions: Not applicable.

Main outcome measure: The KOOS-PF22 (22 items comprising the 2 KOOS domains of activities of daily living and sport/recreation). Its dimensionality was investigated using exploratory factor analysis. RA was performed on KOOS-PF22, KOOS-PS7, and an item set (called KOOS-PS12) obtained by removing from KOOS-PF22 all misfitting items.

Results: KOOS-PF22 showed a 2-factor structure, with the first one largely prevailing. Ten of its 22 items did not fit the Rasch model, while item dependency was present in 7 item pairs. Five of 7 KOOS-PS7 items and all KOOS-PS12 items showed acceptable fit to the Rasch model. The person separation reliability was .79 for KOOS-PS7 and .90 for KOOS-PS12; Cronbach alpha was .81 for KOOS-PS7 and .92 for KOOS-PS12.

Conclusions: RA was not able to replicate the selection process of the items included in KOOS-PS7. KOOS-PS7 reliability levels indicate that the instrument seems more useful for group decisions than for everyday clinical application in single patients. To achieve a short form of KOOS-PF22 valid for a wide spectrum of individuals with knee OA, further psychometric studies are warranted.