Preoperative echocardiography predictive analytics for postinduction hypotension prediction.

Area of Science:

• Cardiology
• Anesthesiology
• Medical Informatics

Background:

• Hypotension is a significant risk factor for adverse perioperative outcomes.
• Preoperative transthoracic echocardiography (TTE) is increasingly used for risk assessment in noncardiac surgery.

Purpose of the Study:

• To develop and evaluate machine learning (ML) models for predicting postinduction hypotension using preoperative TTE data.
• To compare the performance of ML models against conventional statistical models.
• To identify key preoperative echocardiographic factors associated with postinduction hypotension.

Main Methods:

• A retrospective observational study of 1,956 adult patients undergoing general anesthesia.
• Utilized multiple supervised ML classification techniques, including deep neural networks and gradient boosting machines.
• Trained and tested models on preoperative TTE measurements (n=95) to predict postinduction hypotension (MAP <55 mmHg).

Main Results:

• 34% of patients experienced postinduction hypotension.
• The deep neural network model achieved an area under the receiver operating characteristic curve (AUC) of 0.72, outperforming conventional models (AUC ~0.56).
• Important predictors included ascending aorta diameter, transmitral flow A wave, heart rate, and left ventricular mass index.

Conclusions:

• ML models utilizing preoperative TTE data can effectively predict postinduction hypotension.
• These ML models demonstrate higher accuracy than conventional statistical approaches.
• Preoperative echocardiographic parameters are valuable for identifying patients at risk of hypotension during surgery.