Background Frequently used questionnaires to measure fatigue in rheumatoid arthritis (RA) are not developed from the patients’ perspective or have a fixed-length format. Modern psychometrics provide the possibility to measure patient reported outcomes precisely at individual level with few items. To be able to construct a computer adaptive test (CAT) that respectively selects items based on the previous answer of a patient, an item pool has to be constructed and calibrated according to item response theory (IRT).

Objectives Goal of this study was the calibration of an item pool to measure fatigue in RA. It was based on the patients’ perspective captured by an interview study, and examined for face and content validity by a previous Delphi study with patients and professionals. The item pool contained 245 questions about fatigue. The fit of the items with the underlying dimensions was assessed with item response theory (IRT) and the dimensionality structure of the item pool was examined by factor analysis and multidimensional IRT.

Methods Participants were 551 patients with RA from three hospitals in the Netherlands. Obviously, it was not feasible to let each patient score all 245 items of our item pool, so we used an item administration design to construct seven different questionnaire versions. Each patient completed one version of the questionnaire, maximally containing 126 items. IRT analysis using the generalized partial credit model (GPCM) was conducted for each dimension of fatigue. Poorly fitting items were removed. Consecutively, exploratory and confirmatory factor analysis was performed on the remaining items and a multidimensional IRT model was fitted.

Results In the IRT analysis, 49 items showed insufficient item characteristics. Items with a discriminative ability <0.60 and/or model misfit effect sizes >0.10 were removed. Exploratory and confirmatory factor analysis on the 196 remaining items revealed three dimensions of fatigue named: severity, impact and variability of fatigue. The dimensions were further confirmed in multidimensional IRT model analysis.

Conclusions This study provided an initially calibrated multidimensional item bank and has shown which dimensions and items that came forward from previous studies are important for the development of a multidimensional computerized adaptive test (CAT) for fatigue in RA.

Disclosure of Interest None Declared