Reliability and Predictive Validity of a Gait Assessment using Inertial Measurement Units: The Importance of Standardizing Walking Surface and Footwear

ObjectivesEvaluate procedures for analyzing raw accelerometer data (inertial measurement units) to reconstruct the gait cycle using BioKinetoGraph (BKG). We examine whether footwear and walking surface influence gait (BKG) and evaluate test-retest reliability. We also examine the association between BKG and NIH 4-meter gait, and compare BKG to other neurobehavioral measures for predicting concussion symptoms. MethodsIn Study 1, a within-subjects design with 60 participants was used to examine the effects of footwear (shoes/no-shoes) and walking surface (tile floor/grass) on BKG data, and evaluate retest reliability. Study 2 employed a cross-sectional, cohort design of 1,008 participants to assess BKGs correlation with NIH 4-m gait, and prediction of Centers of Disease Control and Prevention (CDC) concussion symptoms relative to previously validated speed and balance measures. Results2x2 ANOVAs illustrate footwear and walking surface effects on BKG for the power, stride, stability, and symmetry, with variable effect sizes. Retest reliability (Pearson rs) for the no shoes/ tile surface condition ranged from .72-.91 (mean = .80, 4-day average interval). BKG correlates significantly with NIH 4-m gait. Regression analyses found BKG predicts CDC concussion symptom endorsement, and outperforms (2-3 fold) BESS and NIH 4-meter gait. ConclusionsGait assessments should be standardized for footwear and especially walking surface. When standardized (no shoes/hard surface) BKG results in strong test-retest reliability. BKG variables are strongly related to NIH 4-m gait, and are superior to standard measures of gait speed and balance when predicting concussion symptoms; offering additional information when predicting the sequalae of concussion. Summary BoxO_ST_ABSWhat is already known on the topic?C_ST_ABSO_LISensor technology to evaluate gait has established reliability and predicts a wide range of medical outcomes. However, the influence of footwear and walking surface on gait has not been studied, nor has a sensor-based gait assessment been compared to conventional measures for predicting concussion symptoms. C_LI What this study adds?O_LIGait sensor data is sensitive to footwear and walking surface, but can produce good reliability when these factors are standardized. Gait sensor scores converge with other validated measures of gait, and sensor-based measures of power, stride, symmetry, and stability can outperform established gait speed and balance measures when predicting CDC concussion symptoms. C_LI How this study might affect research, practice, or policy?O_LIUniform standards for footwear and walking surface are needed when evaluating gait in both research and practice, and sensor-based gait measures can be reliably assessed to provide insight into the behavioral sequelae of concussion, that are superior to simple gait speed. C_LI