Abstract
Ordinal scales with low resolution are used to assess arm function in clinic. These scales may be improved by adding objective kinematic measures. The aim was to analyze within-subject, inter-rater and overall reliability (i.e., including within-subject and inter-rater reliability) and check the system's validity of kinematic measures from inertial sensors for two such protocols on one person. Twenty healthy volunteers repeatedly performed two tasks, finger-to-nose and drinking, during two test sessions with two different raters. Five inertial sensors, on the forearms, upper arms and xiphoid process were used. Comparisons against an optical camera system evaluated the measurement validity. Cycle time, range of motion (ROM) in shoulder and elbow were calculated. Bland⁻Altman plots and linear mixed models including the generalizability (G) coefficient evaluated the reliability of the measures. Within-subject reliability was good to excellent in both tests (G = 0.80⁻0.97) and may serve as a baseline when assessing upper extremities in future patient groups. Overall reliability was acceptable to excellent (G = 0.77⁻0.94) for all parameters except elbow axial rotation in finger-to-nose task and both elbow axial rotation and flexion/extension in drinking task, mainly due to poor inter-rater reliability in these parameters. The low to good reliability for elbow ROM probably relates to high within-subject variability. The sensors provided good to excellent measures of cycle time and shoulder ROM in non-disabled individuals and thus have the potential to improve today's assessment of arm function.