Step-Count Accuracy of 3 Motion Sensors for Older and Frail Medical Inpatients

Objective To measure the step-count accuracy of an ankle-worn accelerometer, a thigh-worn accelerometer, and a pedometer in older and frail inpatients. Design Cross-sectional design study. Setting Research room within a hospital. Participants Convenience sample of inpatients (N=32; age, ≥65 years) who were able to walk 20m independently with or without a walking aid. Interventions Patients completed a 40-minute program of predetermined tasks while wearing the 3 motion sensors simultaneously. Video recording of the procedure provided the criterion measurement of step count. Main Outcome Measures Mean percentage errors were calculated for all tasks, for slow versus fast walkers, for independent walkers versus walking-aid users, and over shorter versus longer distances. The intraclass correlation was calculated, and accuracy was graphically displayed by Bland-Altman plots. Results Thirty-two patients (mean age, 78.1±7.8y) completed the study. Fifteen (47%) were women, and 17 (51%) used walking aids. Their median speed was .46m/s (interquartile range [IQR], .36–.66m/s). The ankle-worn accelerometer overestimated steps (median error, 1% [IQR, −3% to 13%]). The other motion sensors underestimated steps (median error, 40% [IQR, −51% to −35%] and 38% [IQR −93% to −27%], respectively). The ankle-worn accelerometer proved to be more accurate over longer distances (median error, 3% [IQR, 0%–9%]) than over shorter distances (median error, 10% [IQR, −23% to 9%]). Conclusions The ankle-worn accelerometer gave the most accurate step-count measurement and was most accurate over longer distances. Neither of the other motion sensors had acceptable margins of error.