Reliability and validity of a new observation scale to evaluate the upper limb during gait in persons after stroke

Visual gait assessment is a cost-effective and more feasible way to evaluate post-stroke gait deviations in a clinical setting. Most observation scales focus on the lower limb during walking and therefore contain little information concerning the upper limb1,2. However, the upper limbs also contributes to various aspects of functional ambulation3. Therefore, an observation scale was developed to assess the arm swing during walking in persons after stroke. The aim of this study is to examine the inter- and intra-tester reliability and concurrent validity of the upper limb observation scale using two-dimensional (2D) videos of the persons after stroke during walking. Twenty-four persons after stroke (14 female, 10 male; age 54.29 ± 10.9 years, 5.50 ± 29.6 months post-stroke) underwent clinical tests and walked along a 10-meter walkway at self-selected speed. Walking was videotaped (frontal and sagittal view) to score the upper limb observation scale (Fig. 1) afterwards by three different researchers who were blinded from one another. One researcher scored this scale twice with an interval of two weeks. To assess the inter- and intra-tester reliability, intraclass correlation coefficients (ICC), spearman rank correlations (r) and Cronbach’s alpha’s were calculated. Additionally, 3D data, collected from four participants using the Gait Real-time Analysis Interactive Lab (GRAIL, Motek), was compared to the scores on the U.L.O.H.S.W. to validate the 2D observation of the upper limb during walking.Download : Download high-res image (265KB)Download : Download full-size image Inter-tester reliability for the different items varied with ICC’s between 0.254 and 0.885, correlation coefficients (r) between 0.410 and 1.000 (p<0.05, p<0.01) and Cronbach’s alpha between 0.504 and 0.958. For the intra-tester reliability, the ICC’s ranged from 0.594 to 0.957, the correlation coefficients (r) from 0.585 to 0.945 (p<0.01) and the Cronbach’s alpha from 0.738 to 0.978. Scoring the items concerning the more distal parts of the upper limb and the arm swing itself tended to be more reliable compared to the more proximal parts. Percentages of agreement, calculated between the scores on the observation scale and the 3D data to investigate concurrent validity, ranged from 29% (elbow flexion item) to 83% (shoulder abduction item). This is the first study to investigate the inter- and intra-tester reliability and the validity of an observational scale concerning the hemiplegic arm swing during gait. The tool is not yet sufficiently validated as an observation tool of the arm swing during walking in persons after stroke. Scoring the proximal movements of the upper limb appeared to be least reliable. Further research with a larger study population and a renewed version of this scale should provide more information concerning its clinical usability.