Treatment of obstructive sleep apnea is crucial for long term health and reduced economic burden. For those considered for surgery, drug-induced sleep endoscopy (DISE) is a method to characterize location and pattern of sleep-related upper airway collapse. According to the VOTE classification system, four upper airway sites of collapse are characterized: velum (V), oropharynx (O), tongue (T), and epiglottis (E). The degree of obstruction per site is classified as 0 (no obstruction), 1 (partial obstruction), or 2 (complete obstruction). Here we propose a deep learning approach for automatic scoring of VOTE obstruction degrees from DISE videos. We included 281 DISE videos with varying durations (6 s-16 min) from two sleep clinics: Copenhagen University Hospital and Stanford University Hospital. Examinations were split into 5-s clips, each receiving annotations of 0, 1, 2, or X (site not visible) for each site (V, O, T, and E), which was used to train a deep learning model. Predicted VOTE obstruction degrees per examination was obtained by taking the highest predicted degree per site across 5-s clips, which was evaluated against VOTE degrees annotated by surgeons. Mean F1 score of 70% was obtained across all DISE examinations (V: 85%, O: 72%, T: 57%, E: 65%). For each site, sensitivity was highest for degree 2 and lowest for degree 0. No bias in performance was observed between videos from different clinicians/hospitals. This study demonstrates that automating scoring of DISE examinations show high validity and feasibility in degree of upper airway collapse.
Background: Telemedicine has grown significantly in recent years, mainly during the COVID-19 pandemic, and there has been a growing body of literature on the subject. Another topic that merits increased attention is differences in patient and family experience between telehealth and in-person visits. To our team’s knowledge, this is the first study evaluating pediatric and obstetrics outpatients experience with telemedicine and in-person visit types in an academic maternal and children’s hospital, and its correlation with geographic distance from the medical center throughout 2020, during the COVID-19 crisis. Methods: We aim to evaluate and compare patients’ telemedicine and in-person experience for ambulatory encounters based on survey data throughout 2020, during the COVID-19 pandemic, with particular focus on the influence of distance of the patient’s home address from the medical facility. A total of 9,322 patient experience surveys from ambulatory encounters (6,362 in-person and 2,960 telemedicine), in a maternal and children’s hospital during 2020 were included in this study. The percentage of patients who scored the question “Likelihood to recommend practice” with a maximum 5/5 (top box) score was used to evaluate patient experience. The k-means model was used to create distance clusters, and statistical t-tests were conducted to compare mean distances and Top Box values between telemedicine and in-person models. Logistic regression analysis was used to evaluate the correlation between Top Box scores and patients’ distance to the hospital. Results: Top Box likelihood to recommend percentages for in-person and telemedicine were comparable (in-person = 81.21%, telemedicine = 81.70%, p-value = 0.5624). Mean distance from the hospital was greater for telemedicine compared to in-person patients (in-person = 48.89 miles, telemedicine = 61.23 miles, p-value < 0.01). Patients who live farther displayed higher satisfaction scores regardless of the visit type (p-value < 0.01). Conclusions: There is a direct relationship between the family experience and the distance from the considered medical center, during year 2020, i.e., patients who live farther from the hospital record higher Top Box proportion for “Likelihood to Recommend” than patients who live closer to the medical center, regardless of the approach, in-person or telemedicine.
1 June 2020 in Otolaryngologic Clinics of North America
Night-to-night variability (NtNV) in polysomnography (PSG) contributes to diagnostic uncertainty in obstructive sleep apnea (OSA), yet multi-metric evaluations using closely spaced PSG nights-particularly in moderate-to-severe disease-remain limited. The comparative stability of apnea-hypopnea index (AHI) definitions, hypoxic burden (HB), and threshold calibration remains unclear. What is the NtNV across PSG-derived metrics, and how do AHI scoring definitions and threshold calibration influence diagnostic stability in OSA? We performed a retrospective analysis of a prospective study including 147 participants with prior diagnosis or high pretest likelihood of moderate-to-severe OSA who underwent two PSGs within 10 days. NtNV was quantified across 20 PSG-derived metrics. A normalized NtNV matrix was analyzed using PCA followed by unsupervised k-means clustering to identify data-driven variability-pattern groups. Diagnostic stability was compared across AHI definitions (3%/arousal vs. 4%) and HB risk categories. Statistical calibration models derived AHI 4% thresholds aligned with AHI 3%/arousal severity cutpoints. NtNV demonstrated heterogeneity. Maximum heart rate, positional fractions, and sleep latency were most variable, whereas average SpO2, average heart rate, minimum SpO2, and HB were most stable. In the PCA followed by k-means analysis, respiratory event frequency metrics contributed most strongly and separated participants into lower- and higher-respiratory-variability pattern groups. AHI 4% showed higher classification disagreement than AHI 3%/arousal in short-interval (29.9% vs. 21.2% overall; 14.3% vs. 5.4% at the moderate-to-severe threshold) and longitudinal comparisons (45.9% vs. 31.1%; 20.9% vs. 8.2%). HB showed low inter-night disagreement (11.8%). Calibration models aligned AHI 4% thresholds of 6.1-6.9 and 18.4-22.3 events/h with AHI 3%/arousal cutpoints of 15 and 30 events/h. Positional, autonomic, and sleep architecture metrics showed the highest NtNV; respiratory event frequency metrics were intermediate and oxygenation most stable. Greater classification disagreement with AHI 4% was threshold-driven, with implications for hypopnea scoring, and payer policy in OSA diagnosis.