Development and Performance Assessment of Novel Machine Learning Models to Predict Postoperative Pneumonia After Liver Transplantation

Background/Aims: Postoperative pneumonia commonly occurs in patients after orthotopic liver transplantation ( OLT ), contributing to both morbidity and mortality. Until now, it remains a change in predicting postoperative pneumonia among patients following OLT. In this study, we aimed to develop machine learning (ML) models and to assess their performance for predicting postoperative pneumonia in OLT patients. Methods: Preoperative, intraoperative, and postoperative data of 591 adult patients who underwent OLT from January 2015 to September 2019 were retrospectively analyzed in this study. Six ML models, including logistic regression (LR), support vector machine ( SVM ), random forest ( RF ), multilayer perceptron ( MLP ), extreme gradient boosting ( XGBoost ), and gradient boosting machine ( GBM ) were developed using the training set and assessed using the testing set for their performance in predicting operative pneumonia. Findings: Postoperative pneumonia occurred in 42.81% of OLT patients, which contributed significantly to increased postoperative hospital stay and mortality. 14 features, including INR, HCT, PLT, ALB, ALT, FIB, WBC, PT, serum Na+, TBIL, anesthesia time, preoperative length of hospital stay, total fluid transfusion, and operation time, were significantly associated with postoperative pneumonia. Performance comparison of the six ML models revealed that the XGBoost model exhibited the best overall performance in predicting postoperative pneumonia, with the area under the receiver operating characteristics (ROC) curve (AUC) of 0.734, sensitivity of 52.6%, and specificity of 77.5%.  Conclusions: This study has successfully established six novel ML models, of which the XGBoost model has demonstrated best over performance for predicting postoperative pneumonia in OLT patients, and thus it holds promise for clinical application in the future. Trial Registration : No.[2019]02-609-01 Funding Statement: This study was supported by the National Natural Science Foundation of China (Grant No. 81772127; 81974296), Postdoctoral Science Foundation of China (Grant No. 2019M663260) and the Fundamental Research Funds for the Central Universities, China (Grant No. 20ykpy20). Declaration of Interests: The authors declare no conflicts of interest. Ethics Approval Statement: This study was approved by the Ethnic Committee of in the Third Affiliated Hospital of Sun Yat-sen University-Lingnan Hospital (No. [2019]02-609-01). The requirement for informed consent was waived by the committee, mainly due to the retrospective nature of this study