Information on measurement properties of translated versions of the Western Ontario and McMaster Universities (WOMAC) osteoarthritis index is still limited. This study investigated the internal consistency, test-retest reliability/agreement, construct validity, and floor and ceiling effects of Portuguese-Brazil WOMAC applied to civil servants at baseline of ELSA-Brasil Musculoskeletal cohort. Each measurement property was evaluated in the overall sample, in the subgroup reporting knee symptoms, and across different sociodemographic strata (except factorial analyses). Separate analyses were performed for pain, stiffness and function dimensions, considering the knee with the worst score (or right knee if same score in both knees). A total of 1740 participants were included (319 completed WOMAC on 2 occasions), mean age 56.0 (standard deviation = 8.9) years, 46.8% male, 42.1% had knee symptoms. In the overall sample, the range of results for WOMAC's dimensions were: internal consistency = cronbach alpha 0.92-0.98; test-retest reliability = intraclass correlation coefficient 0.85-0.97; standard error of measurement (SEM) = 1.38-5.86; smallest detectable change (SDC) = 3.84-16.25; lowest possible score = 38.8%-61.1% (floor effect present); highest possible score = 0.2%-0.9% (ceiling effect absent). Construct validity was confirmed by hypothesis testing and factorial analysis. Results were similar in the symptomatic group, except for higher SEM and SDC, and the absence of floor effects in pain and function dimensions. Portuguese-Brazil WOMAC showed good overall quality in a nonclinical setting. Variability in measurement properties across different strata of the population should be taken into consideration for the design of future studies using WOMAC.
Keywords: Knee; Measurement properties; Patient-reported outcome measurements; WOMAC.