Estimation of Physical Activity Intensity in Spinal Cord Injury Using a Wrist-Worn ActiGraph Monitor

Abstract

ObjectivesTo derive accelerometer count thresholds for classifying time spent in sedentary, light intensity, and moderate-to-vigorous physical activity (MVPA) in manual wheelchair users (MWUs) with spinal cord injury (SCI). DesignParticipants completed 18 activities of daily living and exercises for 10 minutes each with a 3-minute break between activities while wearing a COSMED K4b2 portable metabolic cart and an ActiGraph activity monitor on the dominant wrist. A linear regression was computed between the wrist acceleration vector magnitude and SCI metabolic equivalent of task (MET) for 80% of the participants to obtain thresholds for classifying different activity intensities, and the obtained thresholds were tested for accuracy on the remaining 20% of participants. This cross-validation process was iterated for 1000 times to evaluate the stability of the thresholds on data corresponding to different proportions of sedentary, light intensity, and MVPA. MET values of 1.5 or lower were classified as sedentary behavior, MET values between 1.5 and 3 were classified as light intensity, and MET values of 3 or higher were classified as MVPA. The final thresholds were then validated on an out-of-sample independent dataset. ParticipantsMWUs (N=17) with SCI in the out-of-sample validation data set. InterventionsNot applicable. SettingResearch lab, community Main Outcome MeasuresAccelerometer thresholds to classify sedentary, light intensity, and MVPA were obtained and their accuracy tested using cross-validation and an out-of-sample dataset. ResultsThe threshold between sedentary and light intensity was 2057 counts-per-minute, and the threshold between light intensity and MVPA was 11,551 counts per minute. Based on the out-of-sample validation, the obtained thresholds had an overall accuracy of 85.6%, with a sensitivity and specificity of 95.3% and 97.4% for sedentary behavior, 87.8% and 84.5% for light intensity, 68.5% and 96.3% for MVPA, respectively. ConclusionAccelerometer-based thresholds can be used to accurately identify sedentary behavior. However, thresholds may not provide accurate estimations of MVPA throughout the day when participants engage in more resistance-based activities.