Auto-CPAP performances to Real-life Leak Patterns: A High-fidelity Modeling Approach

This study explores auto-CPAP devices' response to real-life leak patterns, focusing on their ability to identify respiratory events across different leaks and to accurately estimate leak flow rates. The study used a sophisticated bench model to simulate sleep apnea realistically. Five CPAP devices were tested across respiratory scripts featuring obstructive/central hypopnea/apnea events, across 6 leak patterns. Residual AHI was retrieved from the built’in software. We highlighted a high variability in the devices' performances. Compared to the actual AHI without leaks, the AHI estimation from the built-in software was on average deteriorated by 32.6 ± 19.5% in the presence of intermittent leakage, and by 5.2 ± 7.4% with continuous leakages (p<0.001). Devices struggled to classify the nature and severity of respiratory events accurately, especially under intermittent leak conditions. Some devices showed relatively good accuracy in estimating leak rates, albeit with errors up to 49.3% in certain conditions. This study highlights the complex interplay between leak patterns and auto-CPAP device performance <fig><object-id>erj;64/suppl_68/OA2758/F1</object-id><object-id>F1</object-id><object-id>F1</object-id><graphic></graphic></fig> <fig><object-id>erj;64/suppl_68/OA2758/F2</object-id><object-id>F2</object-id><object-id>F2</object-id><graphic></graphic></fig>