基于决策树法构建非小细胞肺癌患者术后并发乳糜胸风险预测模型

Construction of a Risk Prediction Model for Postoperative Chylothorax in Patients with Non-small Cell Lung Cancer Based on Decision Tree Method

目的探讨非小细胞肺癌(NSCLC)患者术后并发乳糜胸的危险因素,并构建NSCLC患者术后并发乳糜胸的决策树模型。方法选取2018年1月至2021年12月广州市红十字会医院收治的800例行手术治疗的NSCLC患者作为研究对象。运用Logistic回归分析筛选NSCLC患者术后并发乳糜胸的危险因素,采用SPSS Modeler软件构建NSCLC患者术后并发乳糜胸的决策树模型,并分析决策树模型的预测效能。结果800例NSCLC患者中有50例术后并发乳糜胸,发生率为6.25%。Logistic回归分析结果显示,手术方式、淋巴结清扫组数、吸烟史、糖尿病及肥胖是NSCLC患者术后并发乳糜胸的危险因素(OR=2.423,95%CI 1.268~4.631;OR=3.059,95%CI 1.545~6.057;OR=2.827,95%CI 1.407~5.682;OR=2.226,95%CI 1.194~4.151;OR=2.908,95%CI 1.312~6.446)(P<0.05或P<0.01)。NSCLC患者术后并发乳糜胸风险模型选择了肥胖、吸烟史、淋巴结清扫组数、手术方式及糖尿病5个临床特征作为模型的节点,其中肥胖是最重要的预测因子。NSCLC患者术后并发乳糜胸决策树模型的曲线下面积(AUC)为0.752(95%CI 0.676~0.824),Logistic回归模型的AUC为0.741(95%CI 0.669~0.813),两种模型的AUC比较差异无统计学意义(P>0.05)。结论手术方式、淋巴结清扫组数、吸烟史、糖尿病及肥胖等是NSCLC患者术后并发乳糜胸的危险因素,本研究构建的决策树模型可对乳糜胸的风险评估及防治起到积极的帮助作用。

Objective To investigate the risk factors of postoperative chylothorax in patients with non-small cell lung cancer(NSCLC),and to construct a decision tree model for postoperative chylothorax in patients with NSCLC.Methods A total of 800 patients with NSCLC who underwent surgical treatment in Guangzhou Red Cross Hospital from Jan.2018 to Dec.2021 were selected as the research subjects.Logistic regression analysis was used to screen the risk factors of postoperative chylothorax in patients with NSCLC,and SPSS Modeler software was used to construct a decision tree model for postoperative chylothorax in patients with NSCLC,and the predictive performance of the decision tree model was analyzed.Results Among the 800 patients with NSCLC,50 patients developed chylothorax after surgery,and the incidence of chylothorax was 6.25%.Logistic regression analysis showed that surgical methods,number of lymph node dissection groups,smoking history,diabetes and obesity were the risk factors for postoperative chylothorax in patients with NSCLC(OR=2.423,95%CI 1.268-4.631;OR=3.059,95%CI 1.545-6.057;OR=2.827,95%CI 1.407-5.682;OR=2.226,95%CI 1.194-4.151;OR=2.908,95%CI 1.312-6.446)(P<0.05 or P<0.01).In the risk model of postoperative chylothorax in patients with NSCLC,five clinical characteristics including obesity,smoking history,number of lymph node dissection groups,surgical methods,and diabetes were selected as the nodes of the model,and obesity was the most important predictor.The area under curve(AUC)of the decision tree model for postoperative chylothorax in patients with NSCLC was 0.752(95%CI 0.676-0.824),and the AUC of the Logistic regression model for postoperative chylothorax in patients with NSCLC was 0.741(95%CI 0.669-0.813),and there was no significant difference in AUC between the two models(P>0.05).Conclusion Surgical method,number of lymph node dissection groups,smoking history,diabetes and obesity are risk factors for postoperative chylothorax in patients with NSCLC.The decision tree model constructed in this study can positively help the risk assessment and the prevention and treatment of chylothorax.