Abstract
Summary
A single 3D accelerometer is accurate in measuring upper-extremity activity durations, rest periods and intensities, suggesting its feasibility for daily life measurements with patients. Further enhancements are feasible to reduce residual false classifications of intensity from certain activities.
Introduction
Physical activity is an important outcome measure in orthopaedics as it reflects how surgically restored functional capacity is used in daily life. Accelerometer-based activity monitors (AM) are objective, reliable and valid to determine lower extremity activity in orthopaedic patients. However the suitability of a single AM to monitor upper-extremity activity, in terms of quantity and intensity, has not been investigated. This study investigates the suitability and validity of a single AM to measure quantity and intensity of upper-extremity activity.
Method
Five healthy young subjects (25 ± 3 yrs) were included. Subjects underwent a standardised protocol consisting of walking, combing hair, cleaning a desk while standing, brushing teeth and cleaning a window. All one-handed activities were performed with the right arm, as all subjects were right handed. The activities were performed in a fixed order, at self selected speed, for at least 30s. Between the activities, subjects stood still for 10s with their arms next to the body. A light-weight (18g) 3D-accelerometer (f=40Hz) was taped to the right elbow, just above the base of the Hueter triangle, using double sided tape. During the measurement, patients were recorded by a video camera analyzed by an independent human observer as validation reference. AM data and video-recordings were analyzed per second. The time being active (% of time) was determined over the whole measurement and for every activity separately, the percentage of the active time spend in high and low intensity was determined. Video-recording and AM-output were compared by determining Mean Percentage Error (MPE) and the accuracy (100-MPE).
Results
High agreement in measuring upper-extremity (in-) activity was found between AM and video-recordings, showing an accuracy of 93%. Except for walking and combing hair, high agreement between AM and video-recordings was found in measuring activity intensity (accuracy range: 83–100%). 97% of walking was misclassified: video-recordings classify walking as low intensity, the AM as high intensity. Low agreement (58% agreement) for intensity was also found for combing hair.
Discussion and Conclusion
A single AM is accurate in measuring the duration of upper-extremity activity and rest periods in healthy subjects under controlled circumstances. This suggests the suitability of AM to monitor real life upper-extremity activity, which can serve as objective clinical outcome in patients with shoulder complaints. Beyond durations, the AM seems also suitable for measuring activity intensity, showing high accuracy for most activities. The low accuracy in intensity classification of walking and combing hair can be explained by the different interpretations of intensity by observer and AM. For the observer, intensity classification may be force-related (feeling a resistance may refer to high intensity), while intensity classification is acceleration-related for the AM. The false classification of shoulder movement during walking as intense can be resolved by dedicated filters in the detection algorithms. Future algorithms will allow measurement of arm elevations (elbow below/above shoulder) which may be another relevant outcome parameter. However, already this basic AM application validated here may help e.g. in therapeutic decision making, in evaluating therapy effects or providing biofeedback.
