Development and validation of a practical machine learning model to predict sepsis after liver transplantation
Chaojin Chen1, Bingcheng Chen1, Jing Yang1
1Department of Anesthesiology, The Third Affiliated Hospital of Sun Yat-sen University, Guangzhou, People's Republic of China.
Annals of Medicine
|February 15, 2023
Summary
This study developed a machine learning model to predict sepsis after liver transplantation (LT). The Random Forest model identified key variables to aid clinical decisions and improve patient outcomes.
Area of Science:
- Hepatology
- Transplantation Medicine
- Machine Learning in Healthcare
Background:
- Postoperative sepsis is a significant cause of mortality following liver transplantation (LT).
- Early prediction of sepsis is crucial for improving patient outcomes.
Purpose of the Study:
- To develop and validate a machine learning (ML) model for predicting postoperative sepsis within 7 days in liver transplant recipients.
- To identify key pre- and intra-operative variables associated with post-LT sepsis.
Main Methods:
- Retrospective analysis of 677 liver transplant recipients' data.
- Development and evaluation of seven ML models, including Random Forest Classifier (RF).
- Validation of the best-performing model on an independent dataset.
Main Results:
- Sepsis occurred in 31.9% of patients, associated with increased complications, hospital stay, and mortality.
- Key predictors included red blood cell infusion, ascitic fluid removal, blood loss, urine output, anesthesia time, and preoperative total bilirubin.
- The RF model demonstrated strong predictive performance (AUC 0.731 internal, 0.755 external validation).
Conclusions:
- An RF-based predictive model using eight pre- and intra-operative variables was developed.
- This model can assist in clinical decision-making for preventing post-LT sepsis.
- The findings highlight the potential of ML in enhancing patient care after liver transplantation.
