临床肝胆病杂志2025，Vol.41Issue(12)：2545-2552,8.DOI:10.12449/JCH251216
丙型肝炎失代偿期肝硬化患者死亡风险预测模型的建立及评价
Establishment and evaluation of a predictive model for the risk of death in patients with decompensated hepatitis C cirrhosis
摘要
Abstract
Objective To construct a predictive model for the risk of 24-month mortality in patients with hepatitis C-related decompensated liver cirrhosis based on machine learning algorithms,and to compare this model with traditional Child-Pugh score and Model for End-Stage Liver Disease(MELD)score.Methods A total of 490 patients with hepatitis C-related decompensated liver cirrhosis who were hospitalized in The Third People's Hospital of Kunming from January 2022 to April 2024 were enrolled and followed up to December 2024.According to the survival status of the patients during follow-up,they were divided into death group with 81 patients and survival group with 409 patients.Demographic data,comorbidities,and biochemical parameters were collected from all patients.The independent-samples t test or the Mann-Whitney U test was used for comparison of continuous data between two groups,and the chi-square test or the Fisher's exact test was used for comparison of categorical data between groups.The Logistic regression model,the random forest model,and the XGBoost model were used for dataset training,and 10-fold cross validation was performed.The receiver operating characteristic(ROC)curve was plotted,and sensitivity,specificity,area under the ROC curve(AUC),and recall rate were calculated to assess the predictive value of the model.Results Among the 490 patients,there were 339 male patients(69.2%)and 151 female patients(30.8%).There were significant differences between the survival group and the death group in the proportion of patients comorbid with malignant liver tumor,chronic liver failure,hepatic encephalopathy,AIDS or hypocalcemia/hypoproteinemia,as well as the amount of ascites and the proportion of patients without medication(all P<0.05).The assessment of the predictive ability of the three machine learning models showed that the random forest model had the largest AUC of 0.811,which was significantly better than that of the Logistic regression model(0.676)and the XGBoost model(0.798),and based on both AUC and specificity,the random forest model was selected as the optimal predictive model.The variable importance analysis showed that the top 10 variables(i.e.,direct bilirubin,cholinesterase,alpha-fetoprotein,prothrombin time,total bilirubin,high-density lipoprotein cholesterol,alkaline phosphatase,immunoglobulin E,carbohydrate antigen 19-9,and carbohydrate antigen 125)had relatively high contributions to predicting the risk of death.The ROC curve and AUC were used to compare the random forest model with MELD score and Child-Pugh score in terms of their ability to predict the risk of death in patients with hepatitis C-related decompensated liver cirrhosis,and the results showed that the random forest model had had the smallest AUC interval span,suggesting that this model had a significantly better stability than traditional scores.Conclusion Direct bilirubin,cholinesterase,alpha-fetoprotein,prothrombin time,total bilirubin,high-density lipoprotein cholesterol,alkaline phosphatase,immunoglobulin E,carbohydrate antigen 19-9,and carbohydrate antigen 125 are characteristic variables for the risk of 24-month death in patients with hepatitis C-related decompensated liver cirrhosis.The random forest model can significantly improve the predictive efficacy of the risk of death in such patients,with a better performance than traditional Child-Pugh score and MELD score.关键词
丙型肝炎/肝硬化/预后/机器学习Key words
Hepatitis C/Liver Cirrhosis/Prognosis/Machine Learning引用本文复制引用
李娜,刘幸,罗季,李苑莹,朱翔,彭江丽,马国伟,李生浩..丙型肝炎失代偿期肝硬化患者死亡风险预测模型的建立及评价[J].临床肝胆病杂志,2025,41(12):2545-2552,8.基金项目
国家自然科学基金(82260408) National Natural Science Foundation of China(82260408) （82260408）
