Objective: To investigate the dimensionality and construct validity of the Fugl-Meyer Assessment of the upper extremity by using Rasch analysis.
Design: Secondary analysis of pooled data from 2 existing datasets: a randomized therapeutic exercise clinical trial and a cohort longitudinal study of stroke recovery.
Setting: University research center.
Participants: A total of 512 subjects, ages 69.8+/-11.1 years, who were 0 to 145 days poststroke.
Interventions: Not applicable.
Main outcome measures: Dimensionality was examined with principal components analysis and Rasch item-fit statistics. The Rasch-derived item hierarchy was examined for consistency with the expected course of poststroke upper-extremity recovery suggested by the reflex-hierarchical conceptual model underlying the assessment.
Results: Factor loadings and item infit statistics suggested that the 3 reflex items were empirically disconnected from other assessment items. The reflex items were removed. The modified 30-item assessment showed a unidimensional structure. The Rasch-item-difficulty order was not consistent with the expected item order.
Conclusions: The items testing resting-state reflexes may threaten the assessment's dimensionality. With reflex items removed, the assessment is a unidimensional measure of volitional movement. The Rasch-generated item-difficulty order challenges the hierarchical structure implied by the instrument's underlying conceptual framework.