Ventilator variables predicting extubation readiness in extremely premature infants with prolonged mechanical ventilation: A retrospective observational study

Abstract Background The current generation of neonatal ventilators enables periodic storage of set, measured, and calculated ventilatory parameters. Design Retrospective observational study. Objectives To evaluate and identify the ventilatory, demographic, and clinical pre‐extubation variables that are significant for estimating extubation readiness. Methods Eligible subjects included premature infants <33 weeks of gestation weaned from mechanical ventilation (MV) lasting >24 h. A total of 16 relevant ventilator variables, each calculated from 288 data points over 24 h, together with eight demographic and three clinical pre‐extubation variables, were used to create the generalized linear model (GLM) for a binary outcome and the Cox proportional hazards model for time‐to‐event analysis. The achievement of a 120‐h period without reintubation was defined as a successful extubation attempt (EA) within the binary outcome. Results We evaluated 149 EAs in 81 infants with a median (interquartile range) gestational age of 25 +2 (24 +3 –26 +1 ) weeks. Of this, 90 EAs (60%) were successful while 59 (40%) failed. GLM identified dynamic compliance per kilogram, percentage of spontaneous minute ventilation, and postmenstrual age as significant independent positive variables. Conversely, dynamic compliance variability emerged as a significant independent negative variable for extubation success. This model enabled the creation of a probability estimator for extubation success with a good proportion of sensitivity and specificity (80% and 73% for a cut‐off of 60%, respectively). Conclusions Ventilator variables reflecting lung mechanical properties and the ability to spontaneously breathe during MV contribute to better prediction of extubation readiness in extremely premature infants with chronic lung disease.