Gender-, age-, and race/ethnicity-based differential item functioning analysis of the movement disorder society-sponsored revision of the Unified Parkinson's disease rating scale.

OBJECTIVE: Assess MDS-UPDRS items for gender-, age-, and race/ethnicity-based differential item functioning. BACKGROUND: Assessing differential item functioning is a core rating scale validation step. For the MDS-UPDRS, differential item functioning occurs if item-score probability among people with similar levels of parkinsonism differ according to selected covariates (gender, age, race/ethnicity). If the magnitude of differential item functioning is clinically relevant, item-score interpretation must consider influences by these covariates. Differential item functioning can be nonuniform (covariate variably influences an item-score across different levels of parkinsonism) or uniform (covariate influences an item-score consistently over all levels of parkinsonism). METHODS: Using the MDS-UPDRS translation database of more than 5,000 PD patients from 14 languages, we tested gender-, age-, and race/ethnicity-based differential item functioning. To designate an item as having clinically relevant differential item functioning, we required statistical confirmation by 2 independent methods, along with a McFadden pseudo-R2 magnitude statistic greater than "negligible." RESULTS: Most items showed no gender-, age- or race/ethnicity-based differential item functioning. When differential item functioning was identified, the magnitude statistic was always in the "negligible" range, and the scale-level impact was minimal. CONCLUSIONS: The absence of clinically relevant differential item functioning across all items and all parts of the MDS-UPDRS is strong evidence that the scale can be used confidently. As studies of Parkinson's disease increasingly involve multinational efforts and the MDS-UPDRS has several validated non-English translations, the findings support the scale's broad applicability in populations with varying gender, age, and race/ethnicity distributions. © 2016 International Parkinson and Movement Disorder Society.

Duke Scholars

Author Sheng Luo Biostatistics & Bioinformatics, Division of Biostatistics