Automatic detection and monitoring of expiratory flow limitation in non-invasive ventilation

Introduction: Expiratory flow limitation (EFL) in non-invasive ventilation (NIV) can be detected through flow-time curve semiology. Limited non-invasive tools are available for monitoring it. Objectives: To validate an algorithm created for monitoring EFL using mathematical analysis of full night records, based on a mathematical analysis previously validated for isolated respiratory cycles (Lalmolda et al., ERJ Sep 2020, 56 (suppl 64) 1953). Methods: A cross-sectional study included stable patients under chronic NIV. Records from the last 7 days were classified by four NIV experts based on LFE presence. Taking that classification as the reference, the algorithm9s diagnostic accuracy and correlation with spirometry data were assessed. Results: 22 patients (59.1% women), mean age 63.5 years (SD 19.7), were included. Most prevalent diseases were COPD (27.27%) and OHS (22.72%). Analysis of 243 NIV records (61451 minutes) showed LFE in 56.6% (138) according to experts. Diagnosis accuracy of the algorithm is shown in Figure 1. Correlation between the mean LFE-index for each patient and the presence of obstruction in spirometry was: AUC = 0.892; p<0.0001, Cut-off point = 2.678. The FEV1% and the index showed a significant inverse relationship (R = 0.5, R2 = 30%, p<0.05). Conclusions: The developed algorithm reliably detects EFL and shows correlation with the presence and severity of obstruction in spirometry.