Predicting positive airway pressure appointment attendance using machine learning and deep learning: the role of clinical and oximetric indicators
Sleep and Breathing · 2026-05 · 30 (2)
DOI: 10.1007/s11325-026-03660-9 ↗Kurz gefasst
Fachartikel in der Zeitschrift Sleep and Breathing, online erschienen am 02.04.2026. Verfasst von Barış Çi̇l und 2 weiteren Autor:innen. Beteiligte Einrichtung: Department of Pulmonary Diseases, Mardin Training and Research Hospital, Mardin, Türkiye. drbariscil@hotmail.com. Hauptthemen laut MEDLINE: Sleep Apnea, Obstructive, Machine Learning, Deep Learning, Continuous Positive Airway Pressure. Die Arbeit nennt 19 Literaturangaben.
Abstract (Originalsprache)
INTRODUCTION: Participation in positive airway pressure (PAP) therapy is essential for improving long-term health outcomes in individuals diagnosed with obstructive sleep apnea syndrome (OSAS). This study aimed to identify clinical, physiological, and behavioral factors influencing attendance at initial PAP appointments and to develop and compare predictive models using machine learning and deep learning techniques. METHODS: A total of 369 patients with obstructive sleep apnea syndrome (OSAS) were retrospectively analyzed. Among them, 39.2% attended their scheduled PAP therapy appointment and initiated treatment, whereas 60.8% did not. Correlation analyses were performed, and predictive models including logistic regression, decision trees, random forests, support vector machines, and a deep neural network were developed and compared to predict appointment attendance. RESULTS: A high Epworth Sleepiness Scale (ESS) score and greater time spent in the 80–90% nocturnal oxygen saturation range were positively associated with PAP therapy attendance. In contrast, alcohol consumption was a significant behavioral predictor of non-attendance. Hypoxemia-related metrics, such as average nocturnal oxygen saturation, were found to be stronger predictors than the apnea–hypopnea index (AHI). Anthropometric variables—waist circumference, body mass index (BMI), and neck circumference—showed only weak positive correlations with treatment participation. CONCLUSION: This study is among the few to focus on initiation of PAP therapy, rather than long-term adherence, which is the typical focus of most existing research. The machine learning and deep learning models developed here have the potential to support clinical decision-making systems by identifying high-risk individuals early and enabling timely intervention. Integrating these tools into clinical workflows may improve patient engagement and outcomes from the very first step of treatment.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Department of Pulmonary Diseases, Mardin Training and Research Hospital, Mardin, Türkiye. drbariscil@hotmail.com.
- Faculty of Medicine, Department of General Surgery, Department of Artificial Intelligence, Mardin Artuklu University, Mardin, Türkiye.
- Faculty of Medicine, Department of Pulmonary Diseases, Mardin Artuklu University, Mardin, Türkiye.
Zitieren
APA
Çi̇l, B., Baran, M. T., & Kabak, M. (2026). Predicting positive airway pressure appointment attendance using machine learning and deep learning: the role of clinical and oximetric indicators. Sleep and Breathing, 30(2). https://doi.org/10.1007/s11325-026-03660-9
Vancouver
Çi̇l B, Baran MT, Kabak M. Predicting positive airway pressure appointment attendance using machine learning and deep learning: the role of clinical and oximetric indicators. Sleep and Breathing. 2026;30(2). doi:10.1007/s11325-026-03660-9
BibTeX
@article{il2026,
author = {Çi̇l, Barış and Baran, Mehmet Tarık and Kabak, Mehmet},
title = {Predicting positive airway pressure appointment attendance using machine learning and deep learning: the role of clinical and oximetric indicators},
journal = {Sleep and Breathing},
year = {2026},
volume = {30},
number = {2},
doi = {10.1007/s11325-026-03660-9}
}
EUVIMED