Machine Learning-Based Model for Predicting Prolonged Mechanical Ventilation in Patients with Congestive Heart...

Area of Science:

• Cardiology
• Pulmonology
• Artificial Intelligence in Medicine

Background:

• Mechanical ventilation (MV) is crucial for respiratory failure in congestive heart failure (CHF) patients.
• Prolonged MV (PMV) is linked to adverse outcomes, necessitating early identification strategies.

Purpose of the Study:

• To develop and validate a machine learning (ML) model for early prediction of PMV in CHF patients.
• To identify key predictors for PMV in this population.

Main Methods:

• Utilized 12 ML algorithms, with LASSO regression for feature selection from 65 variables.
• Evaluated model performance using Area Under the Curve (AUC) and Brier score.
• Conducted external validation using a separate dataset.

Main Results:

• Identified 10 key features for PMV prediction.
• The CatBoost model demonstrated superior performance (AUC = 0.790) for PMV prediction.
• CatBoost also accurately predicted hospital mortality (AUC = 0.844) and showed generalizability in external validation (AUC = 0.780).

Conclusions:

• A validated CatBoost model accurately predicts PMV in mechanically ventilated CHF patients.
• The model shows promise for improving patient outcomes through early risk stratification.
• The model also effectively predicts hospital mortality in this cohort.