Measures of overnight oxygen saturation to characterize sleep apnea severity and predict postoperative respiratory depression

Abstract

Abstract Background Sleep apnea syndrome, characterized by recurrent cessation (apnea) or reduction (hypopnea) of breathing during sleep, is a major risk factor for postoperative respiratory depression. Challenges in sleep apnea assessment have led to the proposal of alternative metrics derived from oxyhemoglobin saturation (SpO2), such as oxygen desaturation index (ODI) and percentage of cumulative sleep time spent with SpO2 below 90% (CT90), as predictors of postoperative respiratory depression. However, their performance has been limited with area under the curve of 0.60 for ODI and 0.59 for CT90. Our objective was to propose novel features from preoperative overnight SpO2 which are correlated with sleep apnea severity and predictive of postoperative respiratory depression. Methods Preoperative SpO2 signals from 235 surgical patients were retrospectively analyzed to derive seven features to characterize the sleep apnea severity. The features included entropy and standard deviation of SpO2 signal; below average burden characterizing the area under the average SpO2; average, standard deviation, and entropy of desaturation burdens; and overall nocturnal desaturation burden. The association between the extracted features and sleep apnea severity was assessed using Pearson correlation analysis. Logistic regression was employed to evaluate the predictive performance of the features in identifying postoperative respiratory depression. Results Our findings indicated a similar performance of the proposed features to the conventional apnea–hypopnea index (AHI) for assessing sleep apnea severity, with average area under the curve ranging from 0.77 to 0.81. Notably, entropy and standard deviation of overnight SpO2 signal and below average burden showed comparable predictive capability to AHI but with minimal computational requirements and individuals’ burden, making them promising for screening purposes. Our sex-based analysis revealed that compared to entropy and standard deviation, below average burden exhibited higher sensitivity in detecting respiratory depression in women than men. Conclusion This study underscores the potential of preoperative SpO2 features as alternative metrics to AHI in predicting postoperative respiratory.

Keywords

• Sleep apnea
• Oxyhemoglobin saturation (SpO2)
• Postoperative respiratory depression
• Prediction
• Signal processing
• Logistic regression