Predicting Successful Weaning from Mechanical Ventilation by Reduction in Positive End-expiratory Pressure Level...

Area of Science:

• Critical Care Medicine
• Biomedical Informatics
• Machine Learning in Healthcare

Background:

• Mechanical ventilation (MV) weaning is complex, impacting patient outcomes and costs.
• Current weaning methods lack standardization, and prolonged MV is linked to adverse events.
• Predicting weaning readiness is challenging, with positive end-expiratory pressure (PEEP) underutilized as a predictive target.

Purpose of the Study:

• To develop and validate a machine learning model for predicting successful weaning from mechanical ventilation.
• To utilize continuous positive end-expiratory pressure (PEEP) reduction as a target for predicting weaning success.
• To assess the model's performance using established metrics and identify key predictive variables.

Main Methods:

• Retrospective analysis of 12,153 patients from MIMIC-IV and eICU-CRD databases.
• Development of Extreme Gradient Boosting and Logistic Regression models targeting continuous PEEP reduction.
• Model evaluation using AUROC, AUPRC, F1-Score, Recall, PPV, and NPV; variable importance assessed via SHAP.

Main Results:

• The best model achieved an AUROC of 0.84 and AUPRC of 0.69.
• High predictive performance demonstrated with Recall of 0.85, F1-score of 0.86, and PPV of 0.87.
• SHAP analysis identified clinically relevant variables such as MV duration, SaO2, PEEP, and GCS components.

Conclusions:

• Machine learning models show promise in predicting successful mechanical ventilation weaning based on PEEP reduction.
• The developed model, with its high positive predictive value, can assist clinicians in ventilator management decisions.
• This approach offers a novel strategy to optimize weaning protocols and potentially reduce healthcare costs.