Automated Apnea-Hypopnea Index from Oximetry and Spectral Analysis of Cardiopulmonary Coupling

Rationale: The increased prevalence of obstructive sleep apnea (OSA) coincides with a severe shortage of sleep physicians. There is a need for widescale home-sleep-testing devices with accurate automated scoring to accelerate access to treatment.Objectives: To examine the accuracy of an automated apnea-index (AHI) derived from spectral analysis of cardiopulmonary coupling (CPC) extracted from electrocardiograms, combined with oximetry signals, in relation to polysomnograms (PSGs).Methods: Electrocardiograms and pulse-oximeter tracings on PSGs from APPLES (Apnea Positive Pressure Long-term Efficacy Study) were analyzed. Distinct CPC spectral bands were combined with the oxygen desaturation index to create a derived AHI (DAHI). Correlation statistics between the DAHI and the conventionally scored AHI, in which hypopneas required ≥50% airflow reduction alone or a lesser airflow reduction associated with ≥3% desaturation or arousal, using PSGs from APPLES were calculated.Results: A total of 833 adult subjects were included. The DAHI has excellent and strong correlation with the conventionally scored AHI on PSGs, with Pearson coefficients of 0.972 and receiver operating characteristic curves demonstrating strong agreement in all OSA categories: 98.5% in mild OSA (95% confidence interval [CI], 97.6-99.3%), 96.4% in moderate OSA (95% CI, 95.3-97.5%), and 98.5% in severe OSA (95% CI, 97.8-99.2%).Conclusions: An accurate automated AHI can be derived from oximetry and CPC.