Using machine learning to predict perfusionists' critical decision-making during cardiac surgery.

Area of Science:

• Cardiovascular Surgery
• Medical Simulation
• Machine Learning in Healthcare

Background:

• Cardiac surgery operating rooms are complex, high-risk environments requiring expert teamwork.
• Perfusionists are critical in making informed decisions during cardiopulmonary bypass.
• Patient safety and surgical outcomes depend on effective clinical decision-making.

Purpose of the Study:

• To develop predictive models of perfusionist decision-making during critical operating room situations.
• To assess the feasibility of using machine learning for analyzing perfusionist actions.
• To identify potential applications for clinical decision support tools in cardiac surgery.

Main Methods:

• A simulation-based study was conducted using machine learning algorithms.
• 30-fold cross-validation across 30 random seeds was performed.
• Predictive models were trained using data from 148 simulations.

Main Results:

• The machine learning approach achieved 78.2% accuracy in predicting perfusionist actions.
• The 95% confidence interval for accuracy was 77.8% to 78.6%.
• The model demonstrated effectiveness even with a limited number of simulations.

Conclusions:

• Machine learning can effectively model perfusionist decision-making in simulated critical scenarios.
• Findings suggest potential for developing computerized clinical decision support systems for operating rooms.
• These tools could enhance patient safety and improve surgical outcomes in cardiac surgery.