Real-time breath-to-breath asynchrony event detection using time-varying respiratory elastance model

Asynchronous events (AE) occur during mechanical ventilation (MV) therapy when the patient's breathing is not synchronised with the ventilator support. Frequent AE indicates sub-optimal ventilation therapy and may lead to further complications. Asynchrony Index (AI) gives the percentage of AEs as a percentage of total breaths, but is only assessed via manual scrutiny. Thus, there is a need to automate AE detection in real-time. A model-based approach using time-varying elastance to detect AEs is developed and retrospectively assessed in MV patients. Data from 14 mechanically ventilated respiratory failure patients, enrolled in an observational study in Christchurch Hospital, New Zealand were used to investigate the performance of the method. Patient data is sorted according to the ventilation mode used, and AI is calculated for each episode separately. The model-based approach accurately identifies AEs, and shown not to give false positive readings when compared to manual detection (gold standard). None of the ventilation modes give significantly different AI levels (P > 0.05). AI decreases when ventilation mode changes and increases overall time indicate worsen patient-ventilator interaction. The model-based method is able to successfully and accurately calculate AI. Real time use of this metric will enable patients with sub-optimal ventilator settings to be automatically identified for the first time and the settings adjusted as necessary, improving the efficacy of mechanical ventilation therapy, and providing a quantified metric to help guide MV care.