Agreement in Regard to Total Sleep Time During a Nap Obtained Via a Sleep Polygraph and Accelerometer: A Comparison of Different Sensitivity Thresholds of the Accelerometer

Abstract

Appropriate sensitivity threshold of accelerometer to measure total sleep time during nap is not established. Actigraphy-derived total sleep times during naps were calculated using three different sensitivity threshold values and compared with polysomnography. The mean age of the 60 subjects (53 men and 7 women) was 22.8, ranging from 22 to 27years. Determination of the sleep stage by the polygraph and the sleep/wake judgment by the accelerometer obeyed the sleep/wake judgment, and the accelerometer was monitored under different sensitivity threshold settings. The study was carried out during one afternoon with a 3-h nap opportunity. Kappa statistics, correlations, and several indices of accuracy were compared using statistical methods. The mean total sleep times during a nap set for 180min were 160.4, 151.8, and 140.5min, respectively, as judged under the low-sensitive, middle-sensitive, high-sensitive settings of an accelerometer worn on the non-dominant wrist. The corresponding mean total sleep time as calculated using a sleep polygraph was 133.0min. Sleep/wake judgment by three levels of threshold values for the accelerometer showed that high-sensitive threshold showed relatively high specificity (0.452) compared with specificities by the low-sensitive threshold (0.249) or by the middle-sensitive threshold (0.358). The concordance correlation coefficients and 95% confidence intervals (in parenthesis) between the total sleep time judged by polygraph and low-sensitive, middle-sensitive, or high-sensitive accelerometer were 0.40 (0.26-0.51), 0.53 (0.38-0.65), and 0.64 (0.49-0.75), respectively. The Bland-Altman plot of the measurements showed higher agreement between the total sleep time by polygraph and by the accelerometer using the high-sensitive threshold. From the result obtained in this study, the high-sensitive accelerometer showed the strongest agreement of total sleep time and sleep/wake judgment with the calculated value using the sleep polygraph.