ASN's Mission

ASN leads the fight to prevent, treat, and cure kidney diseases throughout the world by educating health professionals and scientists, advancing research and innovation, communicating new knowledge, and advocating for the highest quality care for patients.

learn more

Contact ASN

1401 H St, NW, Ste 900, Washington, DC 20005

email@asn-online.org

202-640-4660

The Latest on Twitter

Kidney Week

Abstract: TH-PO063

Predicting AKI After Cardiac Surgery by Using Machine Learning Methods

Session Information

Category: Acute Kidney Injury

  • 101 AKI: Epidemiology, Risk Factors, and Prevention

Authors

  • Tseng, Po-Yu, Far Eastern Memorial Hospital, New Taipei, Taiwan
  • Yang, David Chih-Yu, Taipei Veterans General Hospital, Taipei, Taiwan
Background

Acute kidney injury (AKI) is an important complication of the cardiac surgeries. Small increases in serum creatinine (SCr) after cardiac surgery have been associated with a significant increase in 30-day mortality. A model that accurately estimates a patient's risk for AKI after cardiac surgery is important in clinical practice. Several risk models have been developed to predict postoperative AKI after cardiac surgery. However, there is less study analyzing clinical big data with the application of machine learning to predict AKI after cardiac surgery.

Methods

We retrospectively enrolled the patients undergoing cardiac surgery (coronary artery bypass graft or valve surgery) in Far East Memorial Hospital from August 2016 to August 2018. The primary outcome was the development AKI.The following machine learning techniques were used: decision tree, random forest, gradient boosting, and support vector machine. The performance of these techniques was compared with that of logistic regression analysis regarding the area under the receiver-operating characteristic curve (AUC). We also used importance matrix plot and shap value to determine the importance of each variables.

Results

A total of 671cases received cardiac surgery. AKI developed in 163(24.3%) patients during the first postoperative week. The highest AUC is 0.829 by the random forest with oversampling .The important matrix plot of random forest revealed that intraoperative urine output, pRBC transfusion during the surgery and and preoperative preoperative serum creatinine were the top three variables contribute to the model.

Conclusion

We successfully use the perioperative parameters to develop the predictive model for AKI after cardiac surgery by using machine learning method.

Comparison of AUC among the different machine learning models

Funding

  • Other NIH Support