Identifying children's nocturnal sleep using 24-h waist accelerometry.

Abstract

The purposes of this study were 1) to add layers and features to a previously published fully automated algorithm designed to identify children's nocturnal sleep and to exclude episodes of nighttime nonwear/wakefulness and potentially misclassified daytime sleep episodes and 2) to validate this refined sleep algorithm (RSA) against sleep logs. Forty-five fourth-grade school children (51% female) participants were asked to log evening bedtime and morning wake time and wear an ActiGraph GT3X+ (ActiGraph LLC, Pensacola, FL) accelerometer at their waist for seven consecutive days. Accelerometers were distributed through a single school participating in the Baton Rouge, USA, site of the International Study of Childhood Obesity, Lifestyle, and the Environment. We compared log-based variables of sleep period time (SPT), bedtime, and wake time to corresponding accelerometer-determined variables of total sleep episode time, sleep onset, and sleep offset estimated with the RSA. In addition, SPT and sleep onset estimated using standard procedures combining sleep logs and accelerometry (Log + Accel) were compared to the RSA-derived values. RSA total sleep episode time (540 ± 36 min) was significantly different from Log SPT (560 ± 24 min), P = 0.003, but not different from Log + Accel SPT (549 ± 24 min), P = 0.15. Significant and moderately high correlations were apparent between RSA-determined variables and those using the other methods (r = 0.61 to 0.74). There were no differences between RSA and Log + Accel estimates of sleep onset (P = 0.15) or RSA sleep offset and log wake time (P = 0.16). The RSA is a refinement of our previous algorithm, allowing researchers who use a 24-h waist-worn accelerometry protocol to distinguish children's nocturnal sleep (including night time wake episodes) from daytime activities.