Comparisons of four methods of estimating physical activity in adult women

PURPOSE: To compare four different methods of measuring physical activity (PA) in adult women under free-living conditions. METHODS: Twelve women participated in a 7-d period during which PA was assessed via self-report, accelerometry and step-counting. Subjects wore at the waist a Tritrac-R3D accelerometer, a Computer Science Application Inc. activity monitor (CSA), both of which measure bodily accelerations in various planes, and a Yamax Digi-Walker-500 that records steps. After the 7-d period subjects responded to a 7-d Physical Activity Recall interview (PAR). RESULTS: Physical activity-related energy expenditure (PAEE) determined from PAR was significantly higher than PAEE estimated from either the Tritrac, CSA, or the Yamax data. Time spent in light, moderate, and hard physical activity was not significantly different between PAR, CSA, and Tritrac. Log-transformed activity counts from CSA were highly correlated with log-transformed Tritrac results (r > 0.90). Number of steps recorded by the Yamax was significantly correlated with untransformed Tritrac and CSA results. CONCLUSIONS: Based upon comparisons with PAR, Tritrac, CSA, and Yamax underestimate the amount of PAEE by 25, 46, and 48%, respectively. The correspondence between the Tritrac and CSA results suggests that both devices produce similar estimates of bodily movement in free-living women. The relationships among results for Tritrac, CSA, and Yamax indicate that the number of steps recorded by the Yamax is representative of the amount of physical activity performed during the day as estimated by Tritrac and CSA. These results should be useful when selecting a method to measure PA in individuals under free-living conditions.