1267 Comparison of a Novel Snoring Application Against In-Laboratory Polysomnography

Abstract Introduction Snoring affects a significant portion of the adult population and can be related to sleep apnea severity. Therefore, measuring snoring intensity may be useful as a tool to predict severity. Many of the currently available applications do not attempt to measure apneas, and only measure snoring, which is only a symptom of OSA. Additionally, the commercial apps have variable strengths and weaknesses for identifying snoring and apneas. The current study aimed to assess a new snoring app, SleepGofer, by comparing snoring and apneas against polysomnography and determining whether the application snoring metrics are predictive of the in-laboratory AHIPSG. Methods Participants (n=74, females=32, Mage= 46.71 years, SD=13.83) were receiving a diagnostic sleep study at the Windsor Sleep Disorders Center. During the PSG, participants also had recordings from SleepGofer. The in-laboratory system consisted of mainly Cadwell sleep systems, using a snore sensor placed on the neck. A composite score measuring snoring severity and estimating apneas was calculated from the application (Ideal SleepSG). Correlations were used to assess the relationship between snoring metrics from the app and PSG. As well, ROC analyses were used to determine the sensitivity and specificity of the application composite score in predicting presence of moderate and severe AHI (AHI>15). Results Ideal SleepSG was negatively associated with the snoring indexPSG (r =- .408, p<.001) and with the AHIPSG (r =- .567, p < .001). AHISG was significantly associated with AHIPSG, (r=.610, p<.001), although AHISG (M=4.32, SE=.830) underestimated events compared to AHIPSG (M=18.02, SE=2.00; F(1,75)=74.47, p<.001). Using Ideal SleepSG to predict presence of moderate and severe OSA, an ROC analysis demonstrated acceptable discriminatory ability, with an AUC of .760 (95% CI: .651-.869). At a threshold value of 140.5 for Ideal SleepSG, (Youden’s index=.432), the model achieved 75.8% sensitivity and 67.4% specificity. With a prevalence of 43.4% of the sample with moderate or severe AHIPSG, the PPV was 64.1% for this cut-off score, with a NPV of 78.4%. Conclusion The novel application, SleepGofer, demonstrated utility as a screening tool for the general population and for identifying individuals at risk for moderate and severe OSA based on snoring intensity. Support (if any) TRS Waterloo Sleep Institute, Canada