Evaluation of internal construct validity and unidimensionality of the Brachial Assessment Tool (BrAT), a patient-reported outcome measure for Brachial Plexus Injury.

Abstract

BACKGROUND: People with Brachial Plexus Injury (BPI) form a very heterogeneous group, with a wide spectrum of ability. While a number of patient-reported outcome measures have been used to assess outcome following adult traumatic BPI, none has been developed or psychometrically evaluated for this population. AIM: To evaluate the internal construct validity and dimensionality of a new patient-reported outcome measure for people with traumatic BPI based on the ICF definition of activity. METHODS: Adults with confirmed traumatic BPI completed a 51 item 5-reponse questionnaire. Responses were analysed in 4 phases: missing responses, item correlations, Exploratory Factor Analysis and Rasch analysis to evaluate the properties of fit to the Rasch model, threshold response, local dependency, dimensionality, Differential Item Functioning and targeting to the sample. RESULTS: 106 adults (age range 18-82) were recruited from five outpatient clinics across Australia. Size items of the 51 items were deleted for missing responses and items rescored to 4 responses. Ten items were deleted for high inter-item correlation >0.81. The remaining 35 items, while demonstrating fit to the Rasch model, showed evidence of local dependency and multidimensionality. Items were divided into three subscales. Following removal of four items all three subscales demonstrated fit to the model with no local dependency, minimal disordered thresholds, no unidimensionality with no DIF for age, time post injury or self-selected dominance. The subscales were combined into 3-testlets and achieved overall fit to the model, no misfit and unidimensionality allowing calculation of a summary score. CONCLUSION: This preliminary analysis of the BrAT supports the internal construct validity of the BrAT a unidimensional targeted 4-response patient-reported outcome measure designed to solely assess activity following traumatic BPI regardless of level of injury, age at recruitment, premorbid limb dominance and time post injury.