Do kinematic measures, added to clinical measures, better predict upper extremity motor impairments at three months post-stroke?

Abstract

ObjectivePredicting post-stroke recovery through prediction models is crucial for choosing appropriate treatment options. However, the existing models predominantly incorporate clinical measures although measurement of movement quality using kinematic measures is essential for distinguishing various types of recovery. Thus, this study aimed at determining if, by considering varied aspects of recovery, adding kinematic measurements over clinical measures would better predict upper extremity (UE) motor impairments at three months post-stroke. Materials and methodsEighty-nine stroke survivors (58.9 ± 11.8 years) were assessed for clinical predictors between 4 and 7 days, kinematic predictors within 1 month, and the impairment outcome of the Fugl Meyer Assessment of the UE (FM-UE) at three months post-stroke. Significant predictors (p<0.05) with a variation inflation factor (VIF) <10 were selected for model development. After performing further step-wise selection, three models incorporating clinical outcomes, kinematic measurements, and a combination of these two, respectively, were formulated. ResultsThe clinical model (R2 = 0.70) included shoulder abduction finger extension (SAFE) scores, the National Institutes of Health Stroke Scale (NIHSS), and the Montreal Cognitive Assessment (MoCA). The kinematic model (R2 = 0.34) included total displacement, total time, and reaction time. The combined model (R2 = 0.72) comprised of SAFE score and shoulder flexion. All the models had a minimal mean squared error on cross validation, which indicated a good validity. ConclusionThe performance of clinical and combined prediction models for predicting three-month post-stroke UE motor recovery was nearly similar. However, in order to detect minimal changes over time and to understand all aspects of motor recovery, there is a need to add instrument-based kinematic measures.