The use of Rasch measurement to improve the Oswestry classification scheme.

Abstract

To use Rasch measurement to assess and modify the original classification categories of the Oswestry Low Back Pain Disability Questionnaire (Oswestry), to examine the hypothesis that the items from the Oswestry form a unidimensional construct and a hierarchical representation of low back pain (LBP) disability, and to compare ordinal Likert resultant scores to interval Rasch scaled scores with disability categories serving as a framework. Rasch analysis model. Existing Oswestry admission data generated by FOTO were analyzed. Statistical analysis of sample database. A sample of 942 patients with LBP referred for physical therapy between 1993 and 1994. Not applicable. Construct validity; disability categories (as assessed by Likert and Rasch models); and fit statistics (infit, outfit as mean squares). All items from the Oswestry except the pain item fit the Rasch model. Construct validity of the scale using the Rasch model required the structure of the rating scale to be modified from 6 response levels to 4. A hierarchical representation of LBP disability was supported. A comparison of the disability categories based on Likert and Rasch scaling revealed them to be nonequivalent. The new scaling changed the disability categories for 44% of patients. Rasch analysis produced disability categories in the Oswestry that are linear and, therefore, useful for quantitatively assessing self-reported disability levels.