Our office will be closed Thursday and Friday, November 23rd-24th for the Thanksgiving holiday. We will reopen at regular business hours on Monday, November 27th.
Purpose: We previously developed and tested the accuracy of an automated algorithm implemented using SAS to detect bed time, wake time, and sleep period time (SPT) in a sample of children. The purpose of this study was to test the accuracy of the algorithm in an adult sample.
Methods: 104 adults were asked to log evening bed time and morning wake time and wear an ActiGraph GT3X+ accelerometer at their waist 24 h/days for seven consecutive days, unless coming in contact with water. Data were averaged for each participant before additional analysis. Mean difference (MD) and mean absolute difference (MAD) were computed. Pearson correlations and dependent sample t-tests were used to compare log-based variables of bed time, wake time, and SPT to corresponding accelerometer determined variables estimated using the automated algorithm.
Results: 85 participants (75% female, BMI=23.9±3.8 kg/m², age=23.3±5.6 years), provided 2+ days of valid accelerometer and log data for a total of 369 days. There was no mean difference between log and accelerometer estimates of bed time (11:48pm vs. 11:53pm, t(84)=1.56, p=.12, MD=5min, MAD=25min, r=.92). However, there was a significant mean difference (t(84)=2.53, p=.01) between log and accelerometer estimates of wake time (7:41am vs. 7:49am, MD=8min, MAD=24min, r=.92). Regardless, there was no mean difference (t(84)=0.70, p=.49) between log accelerometer estimates of SPT (473±59 vs. 476±66 min, MD=3min, MAD=30min, r=.82).
Conclusion: The log and automated accelerometer algorithm estimates were highly correlated and relatively small differences were present. Although a significant mean difference was found for wake time, this difference might not be meaningful in practice (i.e., MD=8 min) and did not ultimately contribute to significant differences in SPT. MAD, which gives a better estimate of the expected differences at the individual level, also provided good validity evidence for the automated algorithm use of individual estimates.