A machine learning-based methodology for stratifying patients into obstructive sleep apnoea risk

Obstructive sleep apnoea (OSA) is a common and chronic disorder that leads to increased day-time sleepiness, is associated with accidents, emerging of cardiovascular and metabolic disorders as well as depression. Almost 20% of the population suffers from OSA while a large portion of people are undiagnosed. The objectives of the current work are: 1) the development of a platform to keep a record of home sleep studies and monitor patients; 2) its use for screening tool for the detection of undiagnosed cases in the general population; 3) the design of a machine learning-based methodology for stratifying patients into high and low risk of OSA based on a series of clinical findings and questionnaires. The proposed methodology showed overall accuracy 87.4%, sensitivity 92.1% and specificity 77.1%.