Rasch analysis of the Burke Lateropulsion Scale (BLS)

Abstract

BACKGROUND: Lateropulsion is a common problem following stroke. Whilst the Burke Lateropulsion Scale (BLS) is recommended in the literature as the outcome measure of choice for measuring lateropulsion, the internal validity of the BLS has not been investigated.

OBJECTIVES: To evaluate the internal validity of the BLS for use in evaluating the effectiveness of therapies aimed at reducing lateropulsion.

METHODS: Rasch analysis procedures were undertaken including assessment of overall model fit, item, and person fit, threshold ordering, differential item functioning, internal consistency, targeting, and dimensionality.

RESULTS: Data from 132 participants were utilized to perform Rasch analysis of the BLS. In this preliminary study, overall model fit and individual item and person fit were found to be good using fit residual statistics and chi-square probability values. The BLS was found to be unidimensional and have good internal consistency (Person Separation Index 0.867). Thresholds for four of the five items were found to be only marginally disordered and were subsequently not modified. Non-uniform differential item functioning was detected for age for the transfers item; however, this item did not display item misfit and was therefore not removed.

CONCLUSIONS: This study identified good psychometric properties of the BLS using Rasch analysis and supports the use of the BLS as a measure of lateropulsion following stroke. Further use of Rasch analysis on the BLS using a larger sample is recommended to confirm these preliminary findings and allow transformation into an interval-level scale.