Diaries Compared To Automated Algorithms And Visual Inspection For Bed Time Estimation At 6-months Postpartum

Abstract

PURPOSE Participant diaries are typically used to identify to-bed time, out-of-bed time, and thus compute total time spent in bed (i.e., ‘bed time’) when using physical activity monitors with 24-hour wear protocols. Automated methods to estimate these transition times and total bed time could reduce researcher and participant burden. This study compared the van Hees and Tracy bed time algorithms and visual inspection to diaries for identification of to-bed time, wake time, and total bed time in postpartum women. METHODS Participants (N = 18) from the PETALS-2 study wore ActiGraph wGT3X-BT devices, initialized at 30 Hz, on the non-dominant wrist for up to 7-days at 6 months postpartum. They logged to-bed and out-of-bed times daily in a paper diary. The van Hees and Tracy algorithms were run on the ActiLife files in R. Visual inspection entailed manual review of the minute-by-minute epoch data. Repeated measures ANOVAs with the Proc Mixed procedure in SAS (to account for repeated data by participant and day of observation, and using an unstructured covariance structure) were used to investigate differences in transition times and total bed time by estimation method. RESULTS Estimated mean to-bed times, out-of-bed times, and total bed times, and their corresponding 95% confidence intervals, are presented in the Table. For total bed time, the van Hees and Tracy algorithms, visual inspection, and diary did not differ from one another (P = .40). For to-bed and out-of-bed times, visual inspection differed from the other methods (P < .0001 for all pair-wise comparisons with visual inspection), but the van Hees and Tracy algorithms did not differ from the diary or each other. CONCLUSION Findings suggest that the van Hees and Tracy algorithms’ estimates of to-bed time, out-of-bed time, and total bed time are comparable to estimates obtained from participant diaries. Replacing participant diaries with these automated methods in large-scale studies is proposed.