Rasch Analysis of the Arabic Fear-Avoidance Beliefs Questionnaire in Individuals with Low Back Pain

Background/Objective: The measurement properties of the Arabic Fear-Avoidance Beliefs Questionnaire (FABQ) have not been examined using the Rasch measurement model. This study evaluated the Physical Activity (FABQ-PA) and Work (FABQ-W) subscales of the Arabic FABQ in individuals with low back pain (LBP). Methods: This cross-sectional study included 113 individuals with LBP who completed the Arabic FABQ. The FABQ-PA and FABQ-W were evaluated separately using RUMM2030. Likelihood-ratio tests supported use of the partial credit model for both subscales. Rasch analysis examined overall and individual item fit, person misfit, response-category threshold ordering, local item dependency, differential item functioning (DIF), person separation, unidimensionality, and targeting. DIF was investigated by sex, age, and LBP duration. Unidimensionality was evaluated by comparing person estimates derived from item subsets defined by principal component analysis of residuals. Targeting was examined using person–item threshold distributions. A previously proposed four-category rescoring structure was additionally explored because of disordered thresholds. Results: Following removal of participants with substantial person misfit, both FABQ-PA (n = 105) and FABQ-W (n = 106) demonstrated satisfactory overall Rasch model fit, and all individual items showed satisfactory fit. Both subscales supported unidimensionality, with no evidence of local item dependency or DIF by sex, age, or LBP duration. Targeting was generally adequate, although coverage was less optimal at the higher end of FABQ-PA and lower end of FABQ-W. Person separation was limited for FABQ-PA (PSI = 0.60) but good for FABQ-W (PSI = 0.80). All items demonstrated disordered thresholds using the original seven-category response scale. A previously proposed four-category rescoring structure improved, but did not completely resolve, threshold disordering. Conclusions: The Arabic FABQ-PA and FABQ-W demonstrated satisfactory final model and item fit and supported unidimensional measurement; however, important limitations were identified. Limited person separation for FABQ-PA and persistent threshold disordering across both subscales indicate that scores should be interpreted cautiously and that further refinement of the FABQ response format is warranted.