Rich Carly, Geraci Marco, Griffiths Lucy, Sera Francesco, Dezateux Carol, Cortina-Borja Mario
Medical Research Centre of Epidemiology for Child Health, University College London, London, United Kingdom.
PLoS One. 2014 Jan 13;9(1):e85134. doi: 10.1371/journal.pone.0085134. eCollection 2014.
Accelerometers are designed to measure plausible human activity, however extremely high count values (EHCV) have been recorded in large-scale studies. Using population data, we develop methodological principles for establishing an EHCV threshold, propose a threshold to define EHCV in the ActiGraph GT1M, determine occurrences of EHCV in a large-scale study, identify device-specific error values, and investigate the influence of varying EHCV thresholds on daily vigorous PA (VPA).
We estimated quantiles to analyse the distribution of all accelerometer positive count values obtained from 9005 seven-year old children participating in the UK Millennium Cohort Study. A threshold to identify EHCV was derived by differentiating the quantile function. Data were screened for device-specific error count values and EHCV, and a sensitivity analysis conducted to compare daily VPA estimates using three approaches to accounting for EHCV.
Using our proposed threshold of ≥ 11,715 counts/minute to identify EHCV, we found that only 0.7% of all non-zero counts measured in MCS children were EHCV; in 99.7% of these children, EHCV comprised < 1% of total non-zero counts. Only 11 MCS children (0.12% of sample) returned accelerometers that contained negative counts; out of 237 such values, 211 counts were equal to -32,768 in one child. The medians of daily minutes spent in VPA obtained without excluding EHCV, and when using a higher threshold (≥19,442 counts/minute) were, respectively, 6.2% and 4.6% higher than when using our threshold (6.5 minutes; p<0.0001).
Quality control processes should be undertaken during accelerometer fieldwork and prior to analysing data to identify monitors recording error values and EHCV. The proposed threshold will improve the validity of VPA estimates in children's studies using the ActiGraph GT1M by ensuring only plausible data are analysed. These methods can be applied to define appropriate EHCV thresholds for different accelerometer models.
加速度计旨在测量合理的人类活动，但在大规模研究中记录到了极高计数值（EHCV）。利用人群数据，我们制定了确定EHCV阈值的方法原则，提出了在ActiGraph GT1M中定义EHCV的阈值，确定了大规模研究中EHCV的出现情况，识别了特定设备的误差值，并研究了不同EHCV阈值对每日剧烈身体活动（VPA）的影响。
我们估计了分位数，以分析从参与英国千禧队列研究的9005名7岁儿童获得的所有加速度计正计数值的分布。通过区分分位数函数得出识别EHCV的阈值。筛选数据以查找特定设备的误差计数值和EHCV，并进行敏感性分析，以比较使用三种处理EHCV方法的每日VPA估计值。
使用我们提出的≥11715次/分钟的阈值来识别EHCV，我们发现，在MCS儿童中测量的所有非零计数值中，只有0.7%是EHCV；在99.7%的这些儿童中，EHCV占总非零计数值的比例小于1%。只有11名MCS儿童（占样本的0.12%）返回的加速度计包含负计数值；在237个这样的值中，有211个计数值在一名儿童中等于-32768。在不排除EHCV的情况下以及使用更高阈值（≥19442次/分钟）时获得的每日VPA分钟数中位数，分别比使用我们的阈值时高6.2%和4.6%（6.5分钟；p<0.0001）。
在加速度计实地工作期间以及在分析数据之前应进行质量控制流程，以识别记录误差值和EHCV的监测器。通过确保仅分析合理的数据，所提出的阈值将提高使用ActiGraph GT1M进行的儿童研究中VPA估计值的有效性。这些方法可用于为不同的加速度计模型定义适当的EHCV阈值。
