Intensity pattern of snoring sounds as a predictor for sleep-disordered breathing

Abstract

The relationship between a new operational definition of sleep sounds and apnea was examined in a population of 69 patients referred for overnight evaluations in a sleep disorders center. The sample contained 18 women (mean age 53.6 years) and 51 men (mean age 48.4 years). Subjects underwent polysomnography (PSG) with concurrent graphical recording of sleep sound intensities throughout the night. An acoustical signature event (ASE) was defined as a loud sound preceded by at least 10 but no more than 90 seconds of silence. Multiple regression was performed using known correlates of apnea and ASE to predict PSG levels of respiratory disturbance. Of the commonly known correlates, only self-reported estimate of snoring and apnea severity explained significant variance to the respiratory disturbance index (RDI; R2 = 0.24, p < 0.0001). ASE was entered into the equation as the last step, significantly improving explained variance (R2Δ = 0.54, p < 0.0001). The final equation R2 was 78% (p < 0.0001). An alternative analysis compared ASE findings to polysomnographic findings in each matched 30-second interval (60,231 observations) in an analysis of receiver's operating characteristics. This analysis resulted in d' = 2.67, indicating acceptable accuracy for screening.