Accuracy Of Three Commonly Used Activity Monitors In Community Dwelling Older Adults

Abstract

Adequate physical activity levels are a key health concern in older individuals. Accurate assessment of physical activity (PA) is necessary in order to identify the sedentary older individual that is in need of intervention. PURPOSE To determine the accuracy of the three objective activity monitors in a group of community dwelling older adults and to explore whether gait speed effects the accuracy of these monitors. METHODS Subjects included 21 community dwelling men and women (mean age 75.9 ± 4.5 years) from suburban Pittsburgh, Pennsylvania. Participants were asked to wear a combination of three activity monitors: the Yamax DigiWalker (DW) pedometer (hip), the ActiGraph (AG) accelerometer (hip), and the StepWatch Activity Monitor (SAM) (ankle). Monitor accuracy was evaluated against observed steps taken during a 100 step walking test. Gait speed was determined with a timed 6 meter walk at the participant's self-selected walking speed. Participants were categorized into three groups based on gait speed (<0.80, 0.80–1.0, > 1.0 m/s). Percent error was calculated as [(monitor steps - observed steps)/observed steps] × 100. RESULTS Preliminary results from 20 participants indicate the DW and AG failed to detect 9.9% and 3.7% of observed steps, respectively and SAM overestimated by 8.2%. Stratified by gait speed, participants with the slowest gait speed (n=3), the DW and AG underestimated by 20.3% and 19.3%, respectively. The SAM overestimated by 10.7%. For those in the middle gait speed category (n=12), the DW and AG again underestimated by 5.5% and 1.7%, respectively. The SAM overestimated by 7.7%. For those with the fastest gait speed (n=5), the DW underestimated by 15%. The AG and SAM overestimated by 0.4% and 8.0%, respectively. Spearman Rank Order Correlation Coefficients indicated a positive association between gait speed and percent error for all three monitors. The AG was the only monitor whose percent error was significantly related to gait speed (rho = 0.47, p = 0.03). CONCLUSIONS When examining the group as a whole, the three activity monitors appear to accurately assess observed steps during a 100 step test. Upon further inspection, in older adults walking slowly, two of the monitors failed to detect 20% of the steps taken. For accurate assessment of PA in older adults, gait speed should be considered. Supported by: University of Pittsburgh, Dept. of Epidemiology Small Grants Fund