0522 Screening and Diagnosis of Obstructive Sleep Apnea in Middle Eastern/North African (MENA) Women: A Pilot Study

Zeina Al Khalil,Hrayr Attarian,Abdulghani Sankari,Emily Herreshoff,Matthew Diemer,Kim Duval,Hadi Ismail,Safa Sankari,Ahmad Sukhon,Ron Chervin,Galit Levi Dunietz
DOI: https://doi.org/10.1093/sleep/zsae067.0522
IF: 6.313
2024-04-20
SLEEP
Abstract:Abstract Introduction The risk of obstructive sleep apnea (OSA) increases with age for both sexes. Beyond chronological age, reproductive aging in women between ages of 40-60 may also raise OSA risk. Associations between reproductive aging and OSA have been infrequently explored among women from underrepresented ethnic groups. Therefore, we initiated a pilot cohort study in women from Middle Eastern/North African (MENA) backgrounds (n=77) from Southeast Michigan. To examine the frequency of OSA, we screened women with the STOP-BMI scoring scheme and then compared their risk profiles against Home Sleep Apnea Test (HSATs). Methods This pilot study included 77 MENA women from Michigan between the ages 40-60y. In addition to demographic, lifestyle and anthropometric data, women were also screened for OSA with the STOP-BMI questionnaire (Score >3 indicating high risk and < 3 indicating low risk). A subset of these women completed HSATs with WatchPAT®️ devices (pAHI 3% >5/hour defined OSA). Results In this sample, the mean age was 48y and a third of women were postmenopausal. Mean BMI was 27.6 and 25% of the women had a BMI>30. Twenty women underwent HSATs, and 15 (75%) showed OSA. The mean BMI of women with and without OSA was 30.6 and 26.8 kg/m2, respectively. Of the 15 women with OSA, eight scored low-risk on the STOP-BMI. Moreover, among women without OSA, 40% had a high-risk score on the STOP-BMI. Data suggested that the STOP-BMI has moderate sensitivity (47%) and specificity (60%). Conclusion We found a high frequency of OSA among midlife women from MENA ancestry and despite a low frequency of obesity. Moreover, the STOP-BMI screening tool for OSA was neither sensitive nor specific among these women. Our data suggest that OSA risk may be independent of BMI and that the current screening tool may not capture the symptoms and risk factors for OSA among midlife women. Support (if any)
neurosciences,clinical neurology
What problem does this paper attempt to address?
The problem that this paper attempts to solve is to evaluate the accuracy and effectiveness of a newly - developed automatic scoring system in identifying and assessing sleep - disordered breathing (SDB), especially for obstructive sleep apnea (OSA). Specifically, the researchers hope to use deep - learning algorithms to automate the processing of polysomnography (PSG) data, thereby reducing the time and effort required for manual scoring and improving the precision of diagnosis. ### Research Background As a traditional diagnostic tool, polysomnography (PSG) has multiple advantages in identifying and assessing various sleep disorders, especially in diagnosing obstructive sleep apnea (OSA). However, the scoring process of PSG data requires a great deal of time and energy. Therefore, the main objective of the research is to evaluate the accuracy of a new automatic scoring system based on deep - learning algorithms in identifying and assessing patients with sleep - disordered breathing (SDB). ### Methods - **Dataset**: A total of 1,000 cases of PSG data were included in the study, of which 700 were used for training, 200 for validation, and 100 (including 25 cases each of simple snoring, mild OSA, moderate OSA, and severe OSA) for testing. - **Model Structure**: This deep - learning model consists of five sequential layers, including an initial perceptron layer, three consecutive long - short - term memory (LSTM) cell layers, and two additional perceptron layers. ### Results - **Patient Characteristics**: The average age of the 100 patients was 50.59 ± 14.01 years, with 55 males and 45 females, and the average body mass index (BMI) was 26.30 ± 3.98. - **Sensitivity and Specificity**: - For identifying apnea or hypopnea events, the sensitivities of the model were 98.06% (96.64 - 99.53), 98.12% (96.68 - 99.87), 98.23% (96.75 - 99.72), and 98.51% (96.2 - 95.91), respectively. - The specificities were 97.79% (92.84 - 99.95), 96.84% (93.6 - 99.95), 96.7% (93.32 - 99.59), and 95.46% (93.71 - 99.08), respectively. - **Area Under the Curve (AUC) for Predicting OSA**: - In the groups with apnea - hypopnea index (AHI) ≥ 5, 15, and 30, the AUC values of the model were 0.9402, 0.9388, and 0.9442, respectively, with no significant differences among the groups. ### Conclusion The research results show that the deep - learning algorithm used exhibits significant accuracy in identifying apnea/hypopnea events and can effectively classify the severity of OSA. ### Support This work was supported by the Technology Development Program (S2718665) of the Ministry of SMEs and Startups in South Korea and the Research Fund of Soonchunhyang University.