Predicting Ground Reaction Force from a Hip-borne Accelerometer during Load Carriage.

INTRODUCTION: Ground reaction forces (GRFs) during load carriage differ from unloaded walking. Methods to quantify peak vertical GRF (pGRFvert) of Soldiers walking with loads outside of a laboratory are needed in order to study GRFs during operationally relevant tasks.

PURPOSE: Develop a statistically based model to predict pGRFvert during loaded walking from ActiGraph GT3X+ activity monitor (AM) vertical acceleration.

METHODS: 15 male Soldiers (25.4±5.3 years, 85.8±9.2 kg, 1.79±9.3 m) wore an ActiGraph GT3X+ AM over their right hip. Six walking trials (0.67-1.58 m/s) with four loads (no load, 15, 27, 46 kg) and two types of footwear (athletic shoes and combat boots) were completed on an instrumented force plate treadmill. Average peak vertical AM acceleration (pACCvert) and pGRFvert were used to develop a regression equation to predict pGRFvert. The model was validated using a leave-one-subject-out (LOSO) approach. Root mean square error (RMSE) and average absolute percent difference (AAPD) between actual and predicted pGRFvert were determined. pGRFvert was also predicted for two novel datasets and AAPD and RMSE calculated.

RESULTS: The final equation to predict pGRFvert included pACCvert, body mass, carried load mass, and pACCvert-carried load mass interaction. Cross-validation resulted in an AAPD of 4.0 ± 2.7% and an RMSE of 69.5 N for LOSO and an AAPD of 5.5 ± 3.9% and an RMSE of 78.7 N for the two novel datasets.

CONCLUSION: A statistically based equation developed to predict pGRFvert from ActiGraph GT3X+ AM acceleration proved to be accurate to within 4% for Soldiers carrying loads while walking. This equation provides a means to predict GRFs without a force plate.