A Machine Learning decision-making tool for extubation in Intensive Care Unit patients.

Area of Science:

• Critical Care Medicine
• Artificial Intelligence in Healthcare
• Predictive Analytics

Background:

• Invasive mechanical ventilation (IMV) is common in critical care.
• Weaning from IMV is challenging, with significant extubation failure rates.
• Accurate prediction of extubation success is crucial for patient outcomes.

Purpose of the Study:

• To develop and evaluate Machine Learning models for predicting extubation outcomes.
• To improve the success rate of weaning from invasive mechanical ventilation.
• To identify predictors of successful extubation using readily available patient data.

Main Methods:

• Utilized a heterogeneous dataset including clinical data, demographics, and respiratory logs.
• Compared three classification algorithms: Logistic Discriminant Analysis, Gradient Boosting, and Support Vector Machines.
• Employed standard preprocessing, hyperparameter tuning, and resampling techniques.

Main Results:

• The Support Vector Machine (SVM) model achieved 94.6% accuracy in predicting extubation outcomes.
• SVM predictors relied on monitor data, medical records, and demographics, unlike Spontaneous Breathing Trials.
• This model offers an alternative to traditional extubation decision criteria.

Conclusions:

• Machine Learning models can accurately predict extubation success in patients on IMV.
• The developed model has the potential to reduce the 9% extubation failure rate.
• Implementing this predictive tool could significantly improve clinical outcomes for critical patients.