From Big Data to AI-Driven Decisions in Obstructive Sleep Apnea: A Narrative Review Integrating the DDPP Framework

Obstructive sleep apnea (OSA) remains underdiagnosed and inadequately managed despite an explosion in multimodal data and swift progress in artificial intelligence (AI). To elucidate the extent of AI techniques utilized in OSA data resources, we conducted a comprehensive literature search in PubMed, Web of Science, Scopus, and IEEE Xplore from 1 April 2020 to 1 April 2025. Search terms related to AI were combined with "obstructive sleep apnea", and 575 original studies were found after de-duplication and exclusion. We employed the DDPP analytics model (Descriptive, Diagnostic, Predictive, and Prescriptive), derived from the business domain, to structure reported clinical applications. The study indicates a significant gap between available data and current AI: most research focuses on sleep monitoring signals, whereas patient-reported outcomes, electronic health records, and environmental data (both social and natural) are largely underutilized. In clinical practice, applications typically concentrate on Descriptive and Diagnostic phases, while Prescriptive analytics for personalized therapy is scarce. This is the first review to assess AI projects from the perspective of OSA data resources, and the first to apply the DDPP framework for sleep medicine analytics. We call on researchers to mine OSA-related data from multiple dimensions and to select suitable AI technologies based on the data characteristics, thereby enhancing clinical decision-making.