Using machine-learning models to predict extubation failure in neonates with bronchopulmonary dysplasia

Area of Science:

• Neonatal Medicine
• Pulmonology
• Artificial Intelligence in Healthcare

Background:

• Bronchopulmonary dysplasia (BPD) is a common complication in preterm infants requiring mechanical ventilation.
• Extubation failure (EF) in neonates with BPD is associated with increased morbidity and mortality.
• Predicting EF is crucial for optimizing respiratory support and patient outcomes.

Purpose of the Study:

• To develop and validate a machine-learning-based decision-support tool for predicting extubation failure in neonates with BPD.
• To identify key clinical factors associated with extubation failure in this population.
• To provide a tool that assists clinicians in determining optimal extubation timing.

Main Methods:

• A dataset of 284 neonates with BPD on mechanical ventilation was analyzed.
• Machine learning algorithms, including extreme gradient boosting (XGBoost), were employed to build predictive models.
• Model performance was evaluated using AUC, DCA, confusion matrices, and feature importance analysis (SHAP values).

Main Results:

• The XGBoost model demonstrated superior performance with an AUC of 0.873, sensitivity of 0.896, and specificity of 0.838.
• Key predictors of EF included pO2, hemoglobin, mechanical ventilation rate, pH, and Apgar scores.
• PO2, hemoglobin, and mechanical ventilation rate were identified as the most significant predictive factors.

Conclusions:

• The XGBoost model is a robust tool for predicting extubation failure in neonates with BPD.
• This predictive model can aid clinicians in making informed decisions regarding extubation timing.
• Early identification of EF risk can help reduce complications and improve outcomes for neonates with BPD.