A Physiology-Informed Prediction Model for Postoperative Respiratory Complications in Obese Patients With Obstructive...

Obstructive sleep apnea (OSA) is a heterogeneous disorder in which impairments in gas exchange and ventilatory control may confer perioperative risk beyond apnea-hypopnea frequency. In individuals with obesity, these physiological vulnerabilities are often compensated during wakefulness but may become clinically manifest under surgical stress, contributing to postoperative respiratory complications (PRCs). Conventional perioperative risk scores rarely incorporate sleep physiology-derived traits. We developed and externally validated a physiology-informed model to predict PRCs in obese patients with OSA undergoing noncardiac surgery. In this prospective single-center cohort study with temporal validation, adults with body mass index≥30 kg/m² and polysomnography-confirmed OSA undergoing noncardiac surgery were enrolled. A physiology-informed prediction model was developed using prespecified clinical variables and sleep physiology markers reflecting nocturnal hypoxemia and ventilatory control limitation. Model performance was assessed using discrimination, calibration, and decision curve analysis. Among 2158 participants, 685 (31.7%) developed PRCs within seven postoperative days. Minimum nocturnal oxygen saturation and preoperative arterial carbon dioxide tension emerged as the most informative physiological predictors, together with selected clinical factors. The model demonstrated good discrimination (area under the curve 0.855 in the derivation cohort and 0.849 in the validation cohort), satisfactory calibration, and consistent net clinical benefit across risk thresholds. These findings suggest that sleep physiology-derived markers of gas-exchange burden and ventilatory control provide prognostic information beyond apnea-hypopnea frequency in obese patients with OSA under perioperative stress and support their integration into perioperative stress-expressed, phenotype-informed risk assessment. Further external validation is warranted before broader clinical application.