To establish the test-retest reliability and concurrent validity of a smartphone-based gait assessment in measuring temporal gait parameters in level-ground walking.
Twenty-six healthy adults (mean age: 20.8±0.7, 13 female) participated in the study. In the first assessment session, they walked back and forth at their comfortable pace on a 15-m pathway for a minimum of 150 steps. An Android smartphone was affixed to the participants’ waist using an elastic belt with the smartphone application to measure the acceleration during the walk. The heel-strike and toe-off were then identified by a customised algorithm using the acceleration data collected. Temporal gait parameters including step time, strike time, single support time, stance phase and swing phase durations were calculated. A motion capture system (Vicon, Centennial, CO, USA) was used to capture body kinematics simultaneously for establishing concurrent validity. Each subject was assessed again after one to four weeks. Pearson Product-Moment Correlation was used to evaluate the concurrent validity of the smartphone and Vicon assessments. Test-retest reliability was examined by the intraclass correlation coefficients (ICC3,1) between measurements from the two sessions.
Step time and strike time, calculated based only on the detection of heel-strike, yielded good test-retest reliability (ICC=0.83-0.85) and excellent concurrent validity (r=0.97-0.98). Duration of sub-phases of a gait cycle, which requires also the detection of toes-off, including single support time, stance phase and swing phase duration, yielded moderate to good test-retest reliability (ICC=0.58-0.80) and fair to moderate concurrent validity (r=0.47-0.70).
Signal analysis showed that our smartphone-based gait assessment, when comparing with Vicon, more accurately detects heel-strike (average error=-0.01±0.03s) than toes-off (average error=-0.05±0.02s). When step time are generated separately for the left and right leg, sub-group analysis indicated that participants with a minimal of 100 valid steps per leg measured yielded better concurrent validity.
Smartphone-based gait assessment has good reliability and validity in assessing step time and stride time. It also allows the delineation of gait sub-phases with fair to moderate validity and moderate to good reliability. Further refinement is required to improve the accuracy of toes-off detection. No less than 100 steps per leg should be collected during clinical application for better validity and reliability.
The result of this study sets the foundation for the development of a smartphone application to facilitate self-assessment of gait quality. The much-improved convenience provides a viable solution for regular mass screening of gait quality in large populations. It has great value in multiple clinical conditions, including fall-risk screening and remote health monitoring.
Technology
Fall
