摘要
目的通过决策树算法对肺癌患者继发肺叶切除术后咳嗽(CAP)构建预测模型。方法回顾性选取2020年1月至2023年7月新余市人民医院和新余市中医院行肺叶切除术的305例肺癌患者作为建模集,根据患者是否继发持续性咳嗽分为CAP组(n=62)和无CAP组(n=243)。另选取2023年8月至2024年10月新余市人民医院和新余市中医院行肺叶切除术的130例肺癌患者作为验证集进行外部验证。收集患者相关资料,对其进行危险因素分析,通过决策树算法构建肺癌患者发生CAP的预测模型。通过受试者工作特征(ROC)曲线评估模型的区分度和预测效能。结果单因素分析显示,两组年龄、吸烟史、手术肺叶、手术范围、周围淋巴结清理情况、术后合并气胸、术后合并胸腔积液、麻醉时间、术后胃酸反流情况比较,差异有统计学意义(P<0.05)。多因素logistic分析结果显示,年龄(β=2.859,OR=17.447,95%CI:3.708~82.092)、手术范围(β=1.814,OR=6.134,95%CI:1.414~26.606)、术后合并胸腔积液(β=-1.891,OR=0.151,95%CI:0.029~0.792)、麻醉时间(β=0.125,OR=1.134,95%CI:1.082~1.188)、术后胃酸反流(β=2.443,OR=11.504,95%CI:2.052~64.491)是肺癌患者CAP的独立影响因素(P<0.05)。决策树算法模型显示,年龄、术后胃酸反流、手术范围、气管周围淋巴结清除、手术肺叶、麻醉时间、术后合并气胸均为CAP的危险因素,而吸烟史为其保护因素。通过绘制风险预测模型的ROC曲线,结果显示预测模型评分曲线约登指数为0.502,AUC为0.783,95%CI:0.707~0.859,敏感度为0.613,特异度为0.889。结论本研究基于决策树算法构建的肺癌患者继发CAP的预测模型具有较好的预测效能,可为临床工作提供参考。
Objective To construct a predictive model for secondary cough after pulmonary lobectomy(CAP)in lung cancer patients through the classification regression tree algorithm.Methods A retrospective selection of 305 lung cancer patients who underwent pulmonary lobectomy in Xinyu People's Hospital and Xinyu Hospital of Traditional Chinese Medicine from January 2020 to July 2023 were made as the modeling set.The patients were divided into the CAP group(n=62)and the non-CAP group(n=243)based on whether they had secondary persistent cough.Another 130 lung cancer patients who underwent pulmonary lobectomy in Xinyu People's Hospital and Xinyu Hospital of Traditional Chinese Medicine from August 2023 to October 2024 were selected as the validation set for external validation.The relevant data of patients were collected,and the risk factors were analyzed.The prediction model of CAP in patients with lung cancer was constructed by classification regression tree algorithm.The discrimination and predictive efficacy of the model were evaluated through the receiver operating characteristic(ROC)curve.Results Univariate analysis showed that there were statistically significant differences between the two groups in terms of age,smoking history,surgical lung lobe,surgical scope,peripheral lymph node clearance,postoperative pneumothorax,postoperative pleural effusion,anesthesia time,and postoperative gastric acid reflux(P<0.05).The results multivariate logistic analysis showed that age(β=2.859,OR=17.447,95%CI:3.708-82.092),surgical range(β=1.814,OR=6.134,95%CI:1.414-26.606),postoperative pleural effusion(β=-1.891,OR=0.151,95%CI:0.029-0.792),anesthesia time(β=0.125,OR=1.134,95%CI:1.082-1.188)and postoperative gastric acid reflux(β=2.443,OR=11.504,95%CI:2.052-64.491)were independent influencing factors for CAP in lung cancer patients(P<0.05).The classification regression tree algorithm model showed that age,postoperative gastric acid reflux,surgical scope,pertracheal lymph node clearance,surgical lung lobe,anesthesia time,and postoperative pneumothorax were all risk factors for CAP,while smoking history was a protective factor for CAP.By drawing the ROC curve of the risk column prediction model,the results showed that the Youden index of the prediction model scoring curve was 0.502,the AUC was 0.783,the 95%CI was 0.707-0.859,the sensitivity was 0.613,and the specificity was 0.889.Conclusion The prediction model for secondary CAP in lung cancer patients constructed based on the decision tree algorithm in this study has good predictive efficacy and can provide a reference for clinical work.
作者
潘丽娟
王玉婷
杨舒媛
刘俊
熊薇
吕红
PAN Lijuan;WANG Yuting;YANG Shuyuan;LIU Jun;XIONG Wei;LYU Hong(Department of Cardiothoracic and Breast Surgery,Xinyu People's Hospital,Jiangxi Province,Xinyu336600,China;Department of Otolaryngology,Xinyu People's Hospital,Jiangxi Province,Xinyu336600,China;Operating Room,Xinyu Hospital of Traditional Chinese Medicine,Jiangxi Province,Xinyu336600,China)
出处
《中国当代医药》
2025年第28期15-20,共6页
China Modern Medicine
基金
江西省卫生健康委科技计划项目(202140623)。
关键词
肺切除术后继发持续性咳嗽
肺癌
决策树
预测模型
Secondary persistent cough after pulmonary lobectomy
Lung cancer
Classification regression tree
Prediction model
