Distinguishing True Sedentary From Accelerometer Non-wearing Time: Accuracy Of Two Automated Wear-time Estimations

作者：ELISABETH A. H. WINKLER, Paul A. Gardiner, Genevieve Nissa Healy, Bronwyn Clark, Takemi Sugiyama, Charles Edwards Matthews, Neville G. Owen · 发表于：Medicine & Science in Sports & Exercise · 年份：2009 · DOI：10.1249/01.mss.0000354172.85505.33 · 被引用次数：10 · 研究领域：Physical Activity and Health、Obesity, Physical Activity, Diet、Mobile Health and mHealth Applications

As large-scale studies of sedentary time (as distinct from not meeting physical activity guidelines) are initiated, the extent to which automated estimations can distinguish non-wear time from periods of sedentary time needs to be examined. Consideration of actual time periods (as distinct from total time) is necessary to accommodate temporal measures, such as bouts or work periods. PURPOSE: To evaluate the accuracy of an existing and a revised wear-time estimation algorithm against self-report behavioral logs. METHODS: A sub-sample of 195 participants in the 2004/05 AusDiab Study wore Actigraph accelerometers (CSA WAM-7164) and were asked to complete logs of accelerometer wear for seven days. Data from 987 observed days with complete, unambiguous log data were analysed using a published method that assigns a non-wear status to any time period of at least 60 minutes duration without three consecutive non-zero counts and with no activity count exceeding 100 (Method 1). We examined a stricter criterion: a 50 count maximum and no more than two non-zero counts in the entire non-wear period (Method 2). RESULTS: The difference between estimated and actual wear minutes was minimal on average, but varied widely (median 0, range -600 to+338). Method 1 misidentified sedentary time as non-wear for 31.9% of observed days; median 78, range 6 to 620 minutes. Method 2 misidentified sedentary time as non-wear for fewer observed days (19.4%); median 75, range 26 to 600 minutes. Both methods w...