A prediction algorithm for severe OSA to facilitate decision for split night study.

BACKGROUND: A split-night sleep study is a specialized diagnostic tool that combines the evaluation and treatment of sleep apnea into a single night. The initial phase involves monitoring physiological parameters to assess the severity of sleep apnea, measured by the Apnea-Hypopnea Index (AHI). Subsequently, Positive Airway Pressure (PAP) therapy is initiated. This study aims to predict full-night AHI and severe obstructive sleep apnea (OSA) based on early-night data. METHODS: Consecutive patients referred for full-night polysomnography (PSG) were included. Clinical data (gender, age, blood pressure, Body Mass Index (BMI), snoring, apnea, neck circumference, ESS) and first 2-hour physiological parameters (sleep amounts, efficiency, respiratory event indices, oxygen saturation) were collected. Multivariate regression analyses were used to predict severe OSA (AHI > 30 events/hr). For external validation, 40 cases suspected of OSA were selected from another sleep center. RESULTS: In a study of 348 patients, 41% were found to have severe OSA. BMI was the most accurate clinical predictor, with an accuracy of 0.69. The initial 2-hour AHI proved to be the best physiological predictor, achieving an accuracy of 0.79. A multivariate model that combined AHI, mean oxygen saturation, and the hypopnea index showed an accuracy of 0.82 for both internal and external samples. CONCLUSIONS: A combined model using various PSG features can accurately identify patients who need split-night PSG. This method could enhance efficiency and lower costs in settings with limited resources.