PROMIS Physical Function Short Forms Display Item- and Scale-Level Characteristics at Least as Good as the Roland Morris Disability Questionnaire in Patients With Chronic Low Back Pain

Alessandro Chiarotto, Leo D. Roorda, Martine H. Crins, Maarten Boers, Raymond W. Ostelo, Caroline B. Terwee

Research output: Contribution to journalArticleAcademicpeer-review

Abstract

Objective: To compare dimensionality, item-level characteristics, scale-level reliability, and construct validity of PROMIS Physical Function short forms (PROMIS-PF) and 24-item Roland Morris Disability Questionnaire (RMDQ-24) in patients with chronic low back pain (LBP). Design: Cross-sectional study. Setting: Secondary care center for rehabilitation and rheumatology. Participants: Patients with nonspecific LBP ≥3 months (N=768). Mean age was 49±13 years, 77% were female, and 54% displayed pain for more than 5 years. Interventions: Not applicable. Main Outcome Measures: Dutch versions of the 4-, 6-, 8-, 10-, and 20-item PROMIS-PF and of the RMDQ-24. Results: PROMIS-PF-6, PROMIS-PF-8, and RMDQ-24 exhibited sufficient unidimensionality (confirmatory factor analysis: comparative fit index>0.950, Tucker-Lewis index>0.950, root means square error of approximation<0.060), whereas the other instruments did not. All instruments were free of local dependence except PROMIS-PF-20 with 4 item pairs with clear residual correlations. Mokken scale analysis found 1 nonmonotone item for PROMIS-PF-20 and 8 for RMDQ-24 (ie, the probability of endorsing these items was not increasing with increasing level on the underlying construct). PROMIS-PF-20 displayed 2 misfitting items (S-χ2 P value>.001). Two-parameter item response theory models found 2 items with low discrimination for RMDQ-24. All other instruments had adequate fit statistics and item parameters. PROMIS-PF-20 displayed the best scale-level reliability. Construct validity was sufficient for all instruments as all hypotheses on expected correlations with other instruments and differences between relevant subgroups were met. Conclusions: PROMIS-PF-6, PROMIS-PF-8, and RMDQ-24 exhibited better unidimensionality, whereas PROMIS-PF-4, PROMIS-PF-6, PROMIS-PF-8, and PROMIS-PF-10 showed superior item-level characteristics. PROMIS-PF-20 was the instrument with the best scale-level reliability. This study warrants assessment of other measurement properties of PROMIS-PF short forms in comparison with disease-specific physical functioning instruments in LBP.
Original languageEnglish
Pages (from-to)297
Number of pages308
JournalArchives of Physical Medicine and Rehabilitation
Volume101
Issue number2
Early online date2 Nov 2019
DOIs
Publication statusPublished - Feb 2020

Cite this