Prediction of prolonged intensive care unit stay after liver transplantation based on machine learning
Xu Jingnan
Li Wenzhe
Wang Jingyan
Wang Jingjie
Zheng Qihang
Yu Xiangyou
Wang Yi
Abstract：Objective To develop a predictive model for prolonged intensive care unit(ICU)stay(pLOS-ICU)in patients after liver transplantation using machine learning techniques.Methods Adult patients who underwent liver transplantation and were admitted to the ICU at the first affiliated hospital of Xinjiang medical university from April 2013 to April 2024 were retrospectively included.Predictive models were established based on ten machine learning.Model performance was evaluated by comparing the area under the receiver operating characteristic curve(AUC)and clinical decision curves.Model interpretability was assessed using Shapley additive explanations(SHAP).Results A total of 230 patients were included and divided into a traning set(n=162)and an internal validation set(n=68).In the training set,from 20 candidate variables encompassing preoperative demographics,underlying diseases,intraoperative data,early postoperative laboratory results,and supportive therapies,seven variables were selected for model construction:the proportion of neutrophils at 8 hours postoperatively,hemoglobin level at 8 hours postoperatively,albumin at 8 hours postoperatively,unconjugated bilirubin at 8 hours postoperatively,alanine aminotransferase(ALT)level at 24 hours postoperatively,whether dopamine was used postoperatively,and whether parenteral nutrition support was used postoperatively.Among all models,the Logistic model performed best(training set AUC=0.797,internal validation set AUC=0.819).SHAP values were used to evaluate feature importance and explain the predictions.Conclusion Early postoperative liver function indicators,inflammatory status,circulatory and nutritional support,and preexisting comorbidities are key predictors of prolonged ICU stay after liver transplantation.SHAP analysis facilitates enables individualized risk assessment and early intervention.
Keywords：Liver transplantationICU length of stayMachine learningRisk prediction modelsShapley additive explanations
Publication Date：2026-01-10
Online Publishing Date：2026-01-29(First online date of this platform, not the publication date of the document)
Pages：9( 21-29 )