Study objective: The goal was to determine the utility and accuracy of automated analysis of single-lead electrocardiogram (ECG) data using two algorithms, cardiopulmonary coupling (CPC), and cyclic variation of heart rate (CVHR) to identify sleep apnea (SA).
Methods: The CPC-CVHR algorithms were applied to identify SA by analyzing ECG from diagnostic polysomnography (PSG) from 47 subjects. The studies were rescored according to updated AASM scoring rules, both manually by a certified technologist and using an FDA-approved automated scoring software, Somnolyzer (Philips Inc., Monroeville, PA). The CPC+CVHR output of Sleep Quality Index (SQI), Sleep Apnea Indicator (SAI), elevated low frequency coupling broadband (eLFCBB) and elevated low frequency coupling narrow-band (eLFCNB) were compared to the manual and automated scoring of apnea hypopnea index (AHI).
Results: A high degree of agreement was noted between the CPC-CVHR against both the manually rescored AHI and the computerized scored AHI to identify patients with moderate and severe sleep apnea (AHI > 15). The combined CPC+CVHR algorithms, when compared to the manually scored PSG output presents sensitivity 89%, specificity 79%, agreement 85%, PPV (positive predictive value) 0.86 and NPV (negative predictive value) 0.83, and substantial Kappa 0.70. Comparing the output of the automated scoring software to the manual scoring demonstrated sensitivity 93%, specificity 79%, agreement 87%, PPV 0.87, NPV 0.88, and substantial Kappa 0.74.
Conclusion: The CPC+CVHR technology performed as accurately as the automated scoring software to identify patients with moderate to severe SA, demonstrating a clinically powerful tool that can be implemented in various clinical settings to identify patients at risk for SA.
Trial registration: NCT01234077.
Keywords: Apnea hypopnea index; Cardiopulmonary coupling; Cyclic variation of heart rate; Sleep apnea.