Item response theory analysis to evaluate reliability and minimal clinically important change of the Roland-Morris Disability Questionnaire in patients with severe disability due to back pain from vertebral compression fractures

Background Context The majority of validation done on the Roland-Morris Disability Questionnaire (RMDQ) has been in patients with mild or moderate disability. There is paucity of research focusing on the psychometric quality of the RMDQ in patients with severe disability. Purpose To evaluate the psychometric quality of the RMDQ in patients with severe disability. Study Design/Setting Observational clinical study. Sample The sample consisted of 214 patients with painful vertebral compression fractures who underwent vertebroplasty or kyphoplasty. Outcome Measures The 23-item version of the RMDQ was completed at two time points: baseline and 30-day postintervention follow-up. Methods With the two-parameter logistic unidimensional item response theory (IRT) analyses, we derived the range of scores that produced reliable measurement and investigated the minimal clinically important difference (MCID). Results Scores for 214 (100%) patients at baseline and 108 (50%) patients at follow-up did not meet the reliability criterion of 0.90 or higher, with the majority of patients having disability due to back pain that was too severe to be reliably measured by the RMDQ. Depending on methodology, MCID estimates ranged from 2 to 8 points and the proportion of patients classified as having experienced meaningful improvement ranged from 26% to 68%. A greater change in score was needed at the extreme ends of the score scale to be classified as having achieved MCID using IRT methods. Conclusions Replacing items measuring moderate disability with items measuring severe disability could yield a version of the RMDQ that better targets patients with severe disability due to back pain. Improved precision in measuring disability would be valuable to clinicians who treat patients with greater functional impairments. Caution is needed when choosing criteria for interpreting meaningful change using the RMDQ.