Sci Rep. 2026 Jul 16. doi: 10.1038/s41598-026-60873-4. Online ahead of print.
ABSTRACT
Monitoring home non-invasive ventilation (NIV) typically relies on commercial ventilator software that processes raw signals to estimate key respiratory parameters. However, the algorithms used by different manufacturers can distort the signals, particularly in the presence of leaks, potentially leading to misdiagnosis. This study compared the performance of built-in software from various ventilators and assessed expert agreement in interpreting processed signals. A lung simulator was used to test ventilators under different conditions: continuous leaks, intermittent leaks, asymmetric inspiratory leaks, circuit disconnection, and upper airway obstruction. A polygraph recorded raw signals for comparison with the software-processed data. Nine NIV experts analyzed 57 randomized ventilator traces, classifying the events without prior knowledge of the simulated conditions. Expert agreement was calculated and statistically analyzed. Expert agreement was high for most scenarios but significantly lower for asymmetric leaks (mean agreement 0.48), which were the main source of error in signal interpretation. Software from the ventilators tended to overestimate tidal volume and misreport leak patterns. Asymmetric leaks caused an upward shift in the baseline flow, mimicking air trapping. No major differences were observed between ventilator manufacturers. The processing of raw signals by built-in software during home NIV can lead to misinterpretations, especially in the presence of asymmetric leaks. These inaccuracies may result in clinical misdiagnoses, such as incorrect detection of asynchrony or air trapping. Improvements in software algorithms are needed to ensure more accurate signal interpretation and to reduce diagnostic errors in home NIV monitoring.
PMID:42457769 | DOI:10.1038/s41598-026-60873-4