2019

Rantalainen, T.; Pirkola, H.; Karavirta, L.; Rantanen, T.; Linnamo, V.

Background. Wearable inertial sensors have been shown to provide valid mean gait characteristics assessments, however, assessment of variability is less convincingly established. Research question. What level of concurrent validity, and session-to-session reliability does an ankle-worn inertial measurement unit (IMU)-based gait assessment with a novel angular velocity-based gait event detection algorithm have among older adults? Methods. Twenty seven (women N = 17) participants volunteered (age 74.4 (SD 4.3) years, body mass 74.5 (12.0) kg, height 165.9 (9.9) cm). Right leg stance, swing, and stride duration and stride length, and stride velocity were concurrently assessed with motion capture and with an IMU from a 3 min self-paced walk up and back a 14 m track repeated twice a week apart. Gait variability was assessed as the SD of all of the registered strides. Results. Significant difference was observed between methods for many of the mean stride characteristics and stride variability (all p < 0.05), fair to excellent agreement was observed for mean values of all of the five stride characteristics evaluated (intra-class correlation coefficient [ICC] from 0.43 to 1.00). However, poor agreement was observed for the SD of all of the evaluated stride characteristics (ICC from -0.25 to 0.00). Both methods indicated excellent session to session reliability for all of the five stride characteristics evaluated (ICC from 0.84 to 0.98, CV%RMS from 1.6% to 3.6%), whereas the variability characteristics exhibited poor to good reliability (ICC from 0.0 to 0.69, CV%RMS from 18.0% to 34.4%). Significance. Excellent concurrent validity and reliability was observed for mean spatiotemporal stride characteristics, however, gait variability exhibited poor concurrent validity and reliability. Although IMUs and the presented algorithm could be used to assess mean spatiotemporal stride characteristics among older individuals, either a more reliable gait event detection algorithm or alternative analytical approaches should be used for gait variability.