肺癌患者合并肺部真菌感染的风险预测模型  

Risk Prediction Model for Pulmonary Fungal Infections in Patients with Lung Cancer

作　　者：杜伟伟 季文涛 罗甜 梁剑平[2] 吕燕华[2] DU Wei-wei;JI Wen-tao;LUO Tian;LIANG Jian-ping;LV Yan-hua(Graduate School of Zunyi Medical University Zhuhai Campus,Zhuhai 519041,China;Department of Respiratory and Critical Care Medicine,Zhongshan City People's Hospital,Zhongshan 528499,China)

机构地区：[1]遵义医科大学珠海校区研究生院,广东珠海519041 [2]中山市人民医院呼吸与危重症医学科,广东中山528499

出　　处：《中山大学学报（医学科学版）》2023年第6期1022-1029,共8页Journal of Sun Yat-Sen University:Medical Sciences

基　　金：国家自然科学基金(82200038)。

摘　　要：【目的】探究肺癌患者合并肺部感染的风险因素,构建和验证一个风险预测模型,使用现有的临床数据来预测肺癌患者的肺部真菌感染风险。【方法】这是一项回顾性研究,收集了2021年1月至2023年3月在中山市人民医院接受治疗的390例肺癌患者的信息,利用合并和不合并肺部真菌感染的肺癌患者人口统计学和临床特征来构建预测发生肺部真菌感染的列线图。所有患者按7:3的比例随机分为训练集和内部验证集两组,应用LASSO回归方法筛选变量和选择预测因子,并使用训练集的多元logistic回归方法构建列线图模型。通过计算受试者工作特征曲线下面积(AUC)确定模型的判断能力,此外,还对模型进行了校正分析和决策曲线分析(DCA)评价预测效果。【结果】LASSO回归筛选出14个潜在的预测因素,进一步的Logistic回归分析结果显示肝损伤、手术、贫血、低蛋白血症、疾病历程、侵入性操作、住院时间大于2周、全身糖皮质激素应用大于2周是肺癌患者发生肺部真菌感染的独立预测因素。根据这些变量建立了一个预测模型,该模型对训练集的AUC95%CI=0.980(0.973,0.896)和内部验证的AUC95%CI=0.956(0.795,1.000),显示具有很高的区分度。训练集和验证集的校准曲线均基本沿45°线分布,DCA曲线显示在阈概率为大于0.03时存在净获益。【结论】肺癌患者合并肺部真菌感染风险预测模型的构建和验证有助于临床确定高危人群,及早进行干预或调整治疗决策。【Objective】To investigate the risk factors for pulmonary fungal infection in lung cancer patients,construct and validate a risk prediction model using available clinical data to predict the risk of pulmonary fungal infections in pa‐tients with lung cancer.【Methods】We conducted a retrospective study and collected information of 390 lung cancer pa‐tients treated at Zhongshan People's Hospital from January 2021 to March 2023.Demographic and clinical characteristics of the patients with and without pulmonary fungal infections were used to construct column line graphs to predict the occur‐rence of pulmonary fungal infections.All enrolled patients were randomly assigned to training set and internal validation set in the ratio of 7:3.For the modelling group,LASSO regression was applied to screen variables and select predictors,and multivariate logistic regression with a training set was used to construct the Noe column line graph model.The judgment ability of the model was determined by calculating the area under the curve(AUC),and in addition,calibration analysis and decision curve analysis(DCA)were performed on the model.【Results】LASSO regression identified 14 potential pre‐dictive factors,and further logistic regression analysis showed that hepatic injury,surgery,anemia,hypoalbuminemia,ill‐ness course,invasive operation,hospital stay at least 2 weeks and glucocorticoid used for at least 2 weeks were indepen‐dent predictors for the occurrence of pulmonary fungal infection in lung cancer patients.A predictive model was established based on these variables,with an AUC95%CI of 0.980(0.973,0.896)for the training set and an AUC95%CI of 0.956(0.795,1.000)for internal validation,indicating high discriminative ability.The calibration curves for both the training set and validation set were distributed along the 45°line,and the decision curve analysis(DCA)showed net benefit for threshold probabilities greater than 0.03.【Conclusions】The construction and validation of a predictive model for the risk

关 键 词：肺癌 肺部真菌感染 危险因素 列线图 模型 

分 类 号：R56[医药卫生—呼吸系统]