Supervised Machine-learning Predictive Analytics for Prediction of Postinduction Hypotension.

Area of Science:

• Anesthesiology
• Machine Learning
• Predictive Analytics

Background:

• Postinduction hypotension is a significant risk factor for adverse perioperative outcomes.
• Machine learning offers robust predictive analytics capabilities by leveraging large datasets.

Purpose of the Study:

• To investigate the efficacy of machine learning methods in predicting the risk of postinduction hypotension.
• To evaluate various supervised machine learning classification techniques for this predictive task.

Main Methods:

• Data from 13,323 patients undergoing general anesthesia were analyzed.
• Supervised machine learning models were applied to predict postinduction hypotension (MAP < 55 mmHg within 10 min).
• Features included preoperative medications, comorbidities, induction agents, and intraoperative vital signs; discrimination was assessed via AUC.

Main Results:

• Postinduction hypotension occurred in 8.9% of cases.
• Gradient Boosting Machine achieved the highest Area Under the Receiver Operating Characteristic curve (AUC) of 0.76 (95% CI, 0.75 to 0.77).
• The best performing model, Gradient Boosting Machine, had a test set AUC of 0.74 (95% CI, 0.72 to 0.77).

Conclusions:

• Machine learning models show feasibility for predictive analytics in anesthesiology.
• Model performance in predicting postinduction hypotension is dependent on model selection and tuning.
• This demonstrates a promising approach for identifying patients at risk of perioperative hypotension.