Estimating physical activity using the CSA accelerometer and a physical activity log.

Purpose: To compare two methods for measuring time spent in physical activity of differing absolute intensities.
Methods: Over a 7-d period, 59 women wore Computer Science and Applications, Inc. (CSA) accelerometers and recorded their activity in physical activity logs (PAL) at 15-min intervals. Three published cut points were used to classify CSA data into resting/light, moderate, and vigorous intensity categories. Data were analyzed using descriptive statistics, Spearman rank-order correlations, and Bland-Altman plots.
Results: The CSA estimates of total (moderate plus vigorous) physical activity using the three cut points ranged from a mean (+/- SD) of 38.1 (+/-26.8) min.d-1 to 312.6 (+/- 101.1) min.d-1. Using the PAL, women self-reported a mean (+/- SD) of 75.1 (+/- 51.7) min.d-1 of total activity. There was fair to modest rank-order agreement between each of the three CSA measures and the PAL measure of total activity, with correlations ranging from r = 0.15 to 0.24. Correlations between CSA and PAL estimates of total activity were higher in women with body mass index values (BMI) below 25 kg.m-2 (r = 0.23-0.38) compared with women with BMI > or = 25 kg.m-2 (r = 0.06-0.08) but did not differ according to age. Correlations between the three CSA cut points ranged from r = 0.45 to 0.86.
Conclusions: Three published cut points designed to classify CSA output by intensity level produced different estimates of physical activity participation. Correlations between CSA and PAL measures of activity intensity were fair overall but higher among leaner women.