Assessing Sensor-Derived Features From a Wrist-Worn Wearable Device as Indicators of Upper Extremity Function in
Tessa C Johnson1, Cole Hagen1, Donna L Coffman2
1Department of Health and Rehabilitation Sciences, College of Public Health, Temple University, Philadelphia, PA.
Objective:
To assess the relationship between sensor-derived features and upper extremity function in individuals with acute and chronic cervical spinal cord injury (cSCI) and to assess the reproducibility of these features in chronic cSCI.
Design:
Prospective, longitudinal study. Participants completed the Capabilities of Upper Extremity Test (CUE-T)-a measure of upper extremity function-at 2 time points, 4 weeks apart, while wearing a wrist-worn inertial measurement unit device on their most-used upper extremity. The device recorded 3-axis accelerometer and gyroscope data from which metrics (features) were derived. Distance correlations (dCorr) assessed associations between features and CUE-T hand and arm function scores. Intraclass correlation coefficients (ICCs) assessed the reproducibility of CUE-T scores and features in the chronic subgroup.
Setting:
Inpatient rehabilitation facility (acute cSCI) and community (chronic cSCI).
Participants:
Forty adults (N=40) with cSCI were enrolled, and 33 provided data.
Interventions:
Not applicable.
Main Outcome Measures:
Correlations between CUE-T scores and features; ICCs of CUE-T scores and features (chronic group).
Results:
At Time 1, 2 features showed strong correlations with CUE-T hand score (dCorr, 0.53-0.58), while 9 showed strong correlations with CUE-T arm score (dCorr, 0.54-0.62). At Time 2, 3 features showed strong correlations with CUE-T hand score (dCorr, 0.53-0.57), while 29 showed strong correlations with CUE-T arm score (dCorr, 0.50-0.72). Types of features were distinct for hand and arm conditions. For the chronic subgroup, CUE-T scores showed excellent reproducibility (ICC, 0.94-0.99), and 16 features demonstrated moderate to good reproducibility (ICC, 0.50-0.77).
Conclusions:
Sensor-derived features can indicate upper extremity function in cSCI, supporting their use for monitoring recovery and functional outcomes. Future research should focus on validating features of upper extremity function to support digital biomarker development and clinical adoption.
