Rasch analysis of the 23-item version of the Roland Morris Disability Questionnaire

Abstract

Objective: To determine the psychometric properties of the 23-item version of the Roland Morris Disability Questionnaire (RMDQ-23) and to quantify their stability across 2 cultures/languages and 2 types of care-settings. Methods: Rasch analysis of data from 1,000 patients with low back pain from primary care (UK and Denmark) and secondary care (Denmark). Results: The RMDQ-23 is unidimensional if local item dependency issues are accommodated, but contains several misfitting or overdiscriminating items, some poor targeting of items, and the scoring of 4-5 items is differentially affected by common clinical characteristics (such as age, gender, pain intensity, pain duration and care setting), depending on the country. Conclusion: As similar results have been found for the RMDQ-24, we believe it is timely to reconsider whether: (i) the RMDQ should be reconstructed using an item-response theory-based approach that includes consideration of new items and response options; or (ii) the use of alternative questionnaires should be recommended, such as the Oswestry Disability Index, that have shown evidence of fitting the Rasch model; or (iii) a completely new condition-specific questionnaire should be developed, perhaps utilizing a computerized adaptive testing platform.