Predicting In-Hospital Acute Kidney Injury after Cardiac Surgery Using Machine Learning.

Area of Science:

• Nephrology
• Cardiology
• Data Science

Background:

• Cardiac surgery-associated acute kidney injury (CSA-AKI) is a significant complication.
• Early prediction of CSA-AKI is crucial for timely intervention and improved patient outcomes.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting CSA-AKI.
• To incorporate clinical variables, biomarkers (BNP, [TIMP-2] × [IGFBP7]), and high-frequency hemodynamic measurements into prediction models.

Main Methods:

• A prospective, observational study involving 602 adult patients undergoing cardiac surgery.
• Trained Random Forest (RF) and Long Short-Term Memory (LSTM) models on perioperative data.
• Evaluated models on a test set of 110 patients to predict CSA-AKI based on creatinine levels.

Main Results:

• The RF model achieved an AUC of 0.73 and accuracy of 0.77 in predicting CSA-AKI.
• The RF model demonstrated a sensitivity of 0.59 and specificity of 0.81.
• The LSTM model achieved an AUC of 0.68 and accuracy of 0.72.

Conclusions:

• Developed two machine learning models incorporating perioperative biomarkers and hemodynamics to predict CSA-AKI.
• Models showed competitive performance, supporting their potential for clinical application.
• These findings lay the groundwork for advancing kidney injury prevention strategies in cardiac surgery patients.