Concurrent Validity of Different Sensor-Based Measures: Activity Counts Do Not Reflect Functional Hand Use in Children and Adolescents With Upper Limb Impairments.

OBJECTIVE: To investigate the concurrent validity of 4 different outcome measures to determine daily functional hand use with wrist-worn inertial sensors in children with upper limb impairments. We hypothesized that the commonly used activity counts are biased by walking and wheeling activities, while measures that exclude arm movements during these periods with activity detection algorithms or by limiting the analysis to a range of functional forearm elevation would lead to more valid estimates of daily hand use. DESIGN: Concurrent validity study with video-based observations of functional hand use serving as the criterion measure. SETTING: The participants were videotaped while performing an activity circuit at the rehabilitation center and wearing inertial sensors. PARTICIPANTS: A convenience sample of 30 school-aged children and adolescents with upper limb impairments. INTERVENTIONS: Not applicable. MAIN OUTCOME MEASURES: Spearman rank correlation coefficients ρ between the criterion measure and 4 sensor-based measures: activity counts, combining activity counts with activity detection algorithms (arm activity counts), limiting activity counts to a functional range of forearm elevation (functional activity counts), and a threshold-based approach limited to the same range of forearm elevation (gross arm movements). RESULTS: Activity counts (ρ=0.43) and gross arm movements (ρ=0.57) did not reveal valid estimates of daily hand use. In contrast, arm and functional activity counts correlated significantly stronger with the criterion measure and revealed valid correlation coefficients of 0.78 and 0.71, respectively. CONCLUSIONS: Activity counts should not be used to measure daily hand use because they are biased by walking and wheeling activities. Arm and functional activity counts provide better and valid alternatives. The selection of these 2 approaches depends on the availability and accuracy of activity detection algorithms and on the users' willingness to wear additional sensors in daily life.