Predicting Postoperative Survival in Patients With Malignant Biliary Obstruction Using an Interpretable Machine Learning Model: A Multicenter Study

利用可解释的机器学习模型预测恶性胆道梗阻患者术后生存率:一项多中心研究

阅读:3

Abstract

BACKGROUND: Endoscopic bile duct drainage is crucial for improving the survival and quality of life in patients with malignant biliary obstruction (MBO). This study aimed to identify key factors that affect postoperative survival in patients with MBO, develop a predictive model, and validate the performance of the model using external data. METHODS: Data were retrospectively collected from The First Hospital of Jiaxing (between 2013 and 2021), The Second Hospital of Jiaxing (between 2014 and 2021), and Affiliated Hospital of Shaanxi University of Traditional Chinese Medicine (between 2014 and 2021). Patient demographics, disease characteristics, and laboratory results of 337 patients were analyzed. Various machine learning models, including the gradient boosted survival tree, extreme gradient boosting (XGBoost), XGBoost accelerated failure time (XGBoost AFT), random survival forests, and Cox proportional hazards regression, were used. The model performance was assessed using the concordance index (C-index), and SHapley Additive exPlanations (SHAP) values were used to interpret the model. RESULTS: The XGBoost AFT model exhibited the best performance, with C-index values of 0.902, 0.722, and 0.705 for the training cohort, test cohort 1, and test cohort 2, respectively. The SHAP analysis revealed that distant metastasis, high total bilirubin level, prolonged prothrombin time, and high-level obstruction significantly impacted survival. A Kaplan-Meier survival analysis demonstrated that the model effectively stratified patients into the high-risk and low-risk groups. CONCLUSIONS: This study provides a robust model that can predict the postoperative survival in patients with MBO. This model was validated using external data, and the results offer valuable guidance for postoperative management and personalized treatment.

特别声明

1、本页面内容包含部分的内容是基于公开信息的合理引用;引用内容仅为补充信息,不代表本站立场。

2、若认为本页面引用内容涉及侵权,请及时与本站联系,我们将第一时间处理。

3、其他媒体/个人如需使用本页面原创内容,需注明“来源:[生知库]”并获得授权;使用引用内容的,需自行联系原作者获得许可。

4、投稿及合作请联系:info@biocloudy.com。