Prediction of Sleep Efficiency and Apnea-Hypopnea Index with Cardiorespiratory Signals Obtained During Sleep-Onset Period

Abstract

This study aimed to propose new effective predictors of sleep parameters using cardiorespiratory signals obtained during the sleep-onset period. First, I attempted to predict sleep efficiency. Sleep efficiency is defined as the ratio between actual sleep time and time spent in bed and the most commonly used measure for the objective assessment of sleep quality. Monitoring sleep efficiency can provide significant information regarding health conditions. It was possible to hypothesize that the autonomic nervous system activity observed before sleep may be associated with sleep efficiency. To assess autonomic activity in the awake resting state, heart rate variability and breathing parameters were analyzed for 5 min. Using these parameters, stepwise multiple linear regression analyses and k-fold cross-validation tests were performed for 240 electrocardiographic and thoracic volume change signal recordings to develop a sleep efficiency prediction model. The models sleep efficiency predictability was evaluated using 60 piezoelectric sensor signal recordings. A regression model was constructed using the power ratio of the low- and high-frequency bands of the heart rate variability signal and the average peak inspiratory flow rate. This model exhibited an absolute error (mean ± SD) of 2.18 ± 1.61% and a Pearsons correlation coefficient of 0.94 (P < 0.01) between the sleep efficiency predictive values and reference values measured with polysomnography. The prediction model has the potential to be utilized for home-based, long-term monitoring of sleep efficiency, and to support effective decision-making regarding the application of sleep efficiency improvement strategies. Second, I attempted to predict the apnea-hypopnea index in obstructive sleep apnea patients. The apnea-hypopnea index is defined as the number of apnea and hypopnea events per hour of sleep and the most widely used quantitative measure for the determination of obstructive sleep apnea severity. With the high prevalence of obstructive sleep apnea, the issue of developing a practical tool for obstructive sleep apnea screening has been raised. Because conventional obstructive sleep apnea screening tools cannot predict the apnea-hypopnea index, their applicability is limited. Here, three predictors of the apnea-hypopnea index were suggested based on the following hypotheses: 1) during the sleep-wake transition period, less irregular respiration cycles would be observed in patients with more severe obstructive sleep apnea, and 2) patients with more severe obstructive sleep apnea would exhibit more attenuated waking vagal tone, which might produce lower effectiveness in decreasing heart rate as a response to deep inspiration breath-holding. Using the three predictors extracted from 120 electrocardiographic recordings, nonlinear regression analyses and k-fold cross-validation tests were performed to develop an apnea-hypopnea index prediction model. For 30 piezoelectric sensor signal recordings, the model exhibited an absolute error (mean ± SD) of 2.66 ± 1.97 events/h and a Pearsons correlation coefficient of 0.99 (P < 0.01) between the apnea-hypopnea index predictive values and reference values measured with polysomnography. The developed apnea-hypopnea index prediction model could be potentially useful to make more reasonable clinical decisions on the need for formal diagnosis and treatment of obstructive sleep apnea.