Predictive Analysis of Extubation Failure in the Paediatric Intensive Care Unit in Bloemfontein, South Africa

Background: Extubation failure (EF) is a significant complication, and it is associated with increased mortality, prolonged hospital stays and extended mechanical ventilation (MV). Determining reliable predictors of EF could improve the clinical decision-making and outcomes. Objectives: Determine the outcomes and predictors of EF in a paediatric intensive care unit (PICU) and develop predictive models using machine learning algorithms. Methods: A retrospective cohort study (n = 824) was conducted in two PICUs in participants who underwent planned extubation (January 2018–December 2022). Demographic characteristics, clinical parameters, ventilator setting, laboratory findings and extubation outcomes were collected. Univariate and multivariate analysis were performed to identify significant predictors of EF. Six machine learning algorithms—Logistic Regression (LR), Artificial Neural Network (ANN), Extreme Gradient Boosting (XGBoost), Random Forest (RF), Support Vector Machine (SVM) and Decision Tree (DT)—were developed and validated for prediction of EF. Results: The overall EF rate was 231 (28%). Multivariate analysis identified a mechanical ventilation for a duration of 3 days or more (aOR 4.49, 95% CI 3.24–6.57, p < 0.001), use of neuromuscular blockade (aOR 1.32, 95% CI 1.07–1.63, p = 0.009), and administration of vasopressors (aOR 1.57, 95% CI 1.24–2.01, p < 0.001) as significant independent predictors of EF. The ANN and LR models demonstrated the highest performance with AUCROC of 0.87 ± 0.04 and 0.86 ± 0.02, respectively. Conclusions: Extubation failure was common in our setting (28%) compared to other studies. Days of ventilation, undernutrition, use of neuromuscular blockade, use of vasopressors or inotropes and CNS comorbidity were associated with EF. The main cause of EF was upper airway obstruction.