1263 Validation of a Novel Snoring Detection Mobile Application Against a Commercially Available Snoring Application

Abstract Introduction Obstructive sleep apnea (OSA) is a common sleep disorder and has a significant public health burden. Snoring is highly prevalent in OSA and is frequently used as a screening criterion before formal assessment. Research has shown that patients generally lack insight into the presence and severity of their own snoring. A phone-based application could allow for the identification of snoring in the community to highlight individuals at risk of OSA. This study aims to validate the performance of a novel AI-based snoring detection application against a commercially available snoring application. Methods Participants (n=76, females=32, Mage= 46.71 years, SD=13.83) were selected from patients receiving a diagnostic sleep study at the Windsor Sleep Disorders Center. During the study, participants had simultaneous recordings of the application being tested, Sleep Gofer (SG), and an already available snoring application, SnoreLab (SL). The proprietary composite scores from each application, based on scoring frequency and intensity, included Ideal Sleep (SG: high values= low snoring) and SnoreScore (SL: high values= high snoring). Correlations were used to assess the relationship between the rankings of snoring severity based on these composite scores, and between the raw composite scores of the two applications. Results The mean Ideal Sleep score was 131.47, SD=39.90, (with a range of 37 to 175), and the mean SnoreScore was 56.38, SD=44.67 (with a range of 0 to 181). These scores were negatively correlated, such that a higher Ideal Sleep Score was significantly related to a lower SnoreScore (r= -.762, p<.001). On average, the absolute difference between rankings of severity was 13.03 ranks (Median=11 ranks), and the rankings of severity based on these snoring applications were positively correlated (r=.711, p<.001). Conclusion The novel application, SleepGofer, demonstrated concurrent validity with SnoreLab when considering composite scores of snoring severity. Generally, the two applications showed alignment on the severity of patient snoring. These results provide initial evidence that the application under development is sensitive to different severities of snoring. Support (if any) A grant by TRS Waterloo Sleep Institute, Canada