Content area
Objective
The purpose of this study is to explore the value of cough sounds and forced exhalation sounds monitored by smartwatches with audio collection capabilities for screening obstructive sleep apnea (OSA) in patients with chronic obstructive pulmonary disease (COPD).
Methods
Stable COPD patients were recruited from an outpatient clinic. All participants completed questionnaires and underwent pulmonary function testing and overnight polysomnography (PSG). A novel smartwatch capable of collecting audio signals was worn to continuously monitor peripheral oxygen saturation (SpO₂), heart rate (HR), heart rate variability (HRV), and respiratory rate (RR). Additionally, voluntary cough and forced exhalation sounds were recorded twice daily. Audio data were denoised, segmented, and analyzed using time- and frequency-domain features. Correlations between audio features and OSA diagnosis/severity were assessed and a predicting model were developed based on these data.
Results
Among the 29 participants with stable COPD, 26 underwent PSG, and 17 were diagnosed with comorbid OSA. Multiple cough and forced exhalation subfeatures correlated significantly with OSA diagnosis and apnea and hypopnea index (AHI). Cough sounds showed the highest correlation with OSA diagnosis ( r = −0.6629, p < 0.001). A logistic regression model using a cough sound subfeature (the median of MFCC_35) achieved 92% accuracy with a Cohen’s kappa value of 0.8276 in predicting OSA in COPD patients.
Conclusion
This study demonstrates a strong association between cough sounds and OSA risk in COPD patients. Cough sounds recorded by smartwatches may serve as a valuable tool for screening OSA in COPD patients, contributing to the management of patients with overlap syndrome.