Machine Learning Models with Preoperative Risk Factors and Intraoperative Hypotension Parameters Predict Mortality...

Area of Science:

• Cardiac Surgery Outcomes
• Machine Learning in Medicine
• Predictive Analytics

Background:

• Postoperative mortality prediction models rarely include intraoperative factors.
• Intraoperative factors like hypotension, vasopressor use, and cardiopulmonary bypass time significantly impact surgical outcomes.

Purpose of the Study:

• To evaluate the predictive ability of machine learning models incorporating intraoperative risk factors for mortality after cardiac surgery.

Main Methods:

• A retrospective study of 5,015 adult cardiac surgery patients (2008-2016).
• Incorporated intraoperative factors (hypotension, vasopressor use, cardiopulmonary bypass time) and preoperative risk factors into five machine learning models: logistic regression, random forests, neural networks, support vector machines, and extreme gradient boosting (XGB).
• Compared model performance using area under the receiver operating characteristic curve (AUC).

Main Results:

• The XGB model, utilizing data from the period outside cardiopulmonary bypass, demonstrated the best predictive performance.
• XGB model achieved an AUC of 0.88 (0.83-0.94), with a specificity of 0.85 (0.83-0.87) and sensitivity of 0.75 (0.57-0.90).

Conclusions:

• The XGB model incorporating intraoperative hypotension outside cardiopulmonary bypass offers superior discrimination and predictive value compared to other models.
• Machine learning models integrating adverse intraoperative factors can enhance risk stratification and patient triaging post-cardiac surgery.