All Times EDT
Keywords: Community mobility, Life Space Assessment Score, Gait analysis, Gait variability, Random Forest Regressor
BACKGROUND Mobility limitations affect about 30 million community-dwelling older adults in the US. It is not understood how gait measured in the lab reflects overall mobility. The Life-Space Assessment score (LSA) provides a summary score of a person’s overall mobility. We determine whether gait speed and LSA are related and which laboratory gait features best describe the LSA. METHODS Gait performance data were collected from 232 participants (age = 77.54 ± 6.56; 60% women; LSA range = 32 to 120, LSA = 74.65 ± 18.57). The two sources of these measures were (1) computerized instrumented walkway for measurement of gait speed, gait variability and walk ratio (2) a tri-axial accelerometer placed at the L3 vertebral level during an over-ground six-minute walking task for measuring gait smoothness, gait regularity, gait symmetry and other time frequency measures. Bivariate correlations of these measures with LSA were computed. Variable selection was performed where multi-collinearities were removed. A random forest model was implemented to get relative variable importance while accounting for known LSA determinants – age, gender, cognition and gait efficacy. RESULTS Gait speed is correlated to LSA (?=+0.26, p<0.01). Low gait variability (step time coefficient of variation, ? =-0.19), higher gait smoothness (harmonic ratios particularly in anterior-posterior direction, ? =0.19), low regularity (entropy rate in vertical direction, ? =-0.18), high gait symmetry (cross correlation between anterior-posterior and vertical direction, ? =0.23), high peak frequency amplitudes (? =0.23) are associated with a high LSA score (p<0.05). Relative variable importance given by a random forest regressor (Test set RMSE = 19.50) shows gait measures are important determinants of LSA. CONCLUSION Analysis of gait variability, gait smoothness, gait symmetry and other time-frequency trunk acceleration measures can inform clinical interventions to improve walking and community mobility.