fetching data ... 
Background: Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) is commonly used in research studies of knee osteoarthritis (OA). It is commonly reported as ordinal raw scores. Previous validation studies using Rasch models requires item deletion. Conversion table to Interval scale for meaningful measurement is not available.
Objectives: We aim to evaluate the fit of WOMAC to the Rasch model in a large sample of patients with knee OA who have undergone knee arthroplasty (KA) to derive the transformation table for interval scale measurement.
Methods: We used data collected from knee OA patients undergoing KA before surgery and at 6-months after KA. We collected demographics, WOMAC, Short-form-36 (SF-36) and Oxford Knee Score (OKS) with the latter two as comparison. WOMAC was calibrated for fit to Rasch model for monotonicity, homogeneity, local item independence and absence of differential item functioning (DIF) in a randomly selected sample of 300 patients from each time point. Parameter estimates were then imported into the full data set to provide the necessary estimates of all the scales. We report responsiveness through Standard Error of Measurement (SEM); Smallest Detectable Difference (SDD), %SDD and effect sizes (ES) between pre-operative and 6-months. Transformation tables of the ordinal to interval values for WOMAC was performed.
Results: Data was analysed for 1136 patients (mean age 65.9 years, 69% female) who underwent KA. WOMAC pain, function and total scores had adequate fit to unidimensional Rasch model with good reliability (Person Separation Index: 0.76, 0.80 and 0.79). No item deletion was required. The SEM, SDD and %SDD of WOMAC total were 4.4, 6.9, 10.1, and ES was 1.97 between pre-operative and 6-months. No significant DIF was seen for age, sex, body mass index, education level or uni-compartmental versus total KA. WOMAC pain, function and total scales were transformed to interval scales.
Conclusion: The WOMAC total, pain and function scales had adequate fit to the Rasch model, providing unidimensional measure with good reliability and responsiveness. Transformation of WOMAC to interval scale measurement is feasible and applicable to wide ranges of studies.
Fit of Scales of WOMAC to the Rasch Model
| WOMAC/ subscale | Residuals | Conditional Chi-Square | Reliability | Unidimens-ionality | ECV | Responsiveness (Interval Scale) | |||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Item | Person | Value(df) | P | PSI | Α | % Significant t-tests | SEM | SDD | ES(6m-baseline) | ||
| Pain | 2.499 | 0.779 | 18.7(10) | 0.044 | 0.76 | 0.87 | 3.03 | 1.0 | 2.9 | 1.04 | 2.11 |
| Stiffness | 1.215 | 0.462 | 57.0(9) | <0.001 | 0.60 | 0.82 | 2.53 | - | - | - | - |
| Functioning | 2.496 | 0.702 | 47.0(42) | 0.274 | 0.80 | 0.94 | 1.16 | 0.93 | 10.2 | 2.49 | 1.91 |
| Total | 4.193 | 0.699 | 43.3(58) | 0.925 | 0.80 | 0.80 | 2.03 | 0.90 | 9.9 | 4.4 | 1.97 |
Acknowledgements: This study is supported by Singapore National Medical Research Council (NMRC/HSRG/0061/2016 and NMRC/CSAINV/022/2017).
Disclosure of Interests: Ying Ying Leung Speakers bureau: Received speaker fee from AbbVie, DKSH, Janssen, Novartis and Pfizer., Julian Thumboo: None declared, Seng Jin Yeo: None declared, Vikki Wylde: None declared, Alan Tennant: None declared