Development and Evaluation of New Accelerometer Cut Points for Adolescent Girls

Thumbnail Image


Kuo_umd_0117E_10655.pdf (1.92 MB)
No. of downloads: 1146

Publication or External Link






Current negative trends in adolescent disease risk factors (e.g., overweight) may be related to physical activity. To study these relationships using accelerometers, how to estimate physical activity from accelerometer counts must be better understood. PURPOSES: (1) To develop new accelerometer cut points for estimating physical activity using disease risk factors as criteria. (2) To evaluate how estimates of physical activity using these newly developed cut points agree with comparison measures (i.e., a previously suggested cut point and self-report physical activity recall questionnaires). METHODS: National Health and Nutrition Examination Survey (NHANES) and Trial of Activity for Adolescent Girls (TAAG) data were examined. New cut points were developed using iterative correlations and signal detection and receiver operating characteristic (ROC) curves. To identify new cut points, potential cut points were identified in a development sample and validated in an evaluation sample. Agreement between new cut points and comparison measures was examined using concordance correlation coefficients, Bland-Altman plots, McNemar's tests, and proportions of agreement. RESULTS: Using the correlation method, two new combinations of light, moderate, and vigorous intensity cut points were identified in NHANES (1900, 4300, and 10000 counts/min and 1900, 4000, and 5000 counts/min) and two in TAAG (1450, 1950, and 2450 counts/30 sec and 1050, 1550, and 2050 counts/30 sec). Using the signal detection/ROC curve method, eleven new cut points were identified in NHANES (ranging from 100 to 2300 counts/min) and three in TAAG (ranging from 100 to 200 counts/min). Concordance correlation coefficients for minutes of activity with a previously suggested cut point tended to be stronger (≥ 0.60) with higher cut points (≥ 2300 count/min), while those with questionnaires were less than 0.10 or the 95% confidence intervals included zero. One new cut point (1800 counts/min) was similar (p = 0.6) to a comparison measure for classifying meeting recommendations. CONCLUSIONS: Some cut points may be more strongly associated with disease risk factors than previously suggested cut points developed using oxygen consumption, but associations are not strong. The new cut points and comparison measures may be measuring different aspects of physical activity, as they were in poor agreement.