感染患者急性肾损伤临床预测模型的构建及验证  

The Construction and Validation of A Clinical Prediction Model for Acute Kidney Injury in Patients withInfectious Diseases

作　　者：靳敏燕 胡渊龙 高超[2] 张倩[2] 邱占军[2] JIN Min-Yan;HU Yuan-Long;GAO Chao;ZHANG Qian;QIU Zhan-Jun(First School of Clinical Medicine,Shandong University of Traditional Chinese Medicine,Jinan 250014,China;Department of Respiratory and Critical Care Medicine,Afiliated Hospital of Shandong University of Traditional Chinese Medicine,Jinan 250014,China)

机构地区：[1]山东中医药大学第一临床医学院,济南250014 [2]山东中医药大学附属医院呼吸与危重症医学科,济南250014

出　　处：《中国药物经济学》2023年第8期34-38,45,共6页China Journal of Pharmaceutical Economics

基　　金：山东省医药卫生科技发展计划(2018WS193)。

摘　　要：目的构建感染患者急性肾损伤(AKI)临床预测模型并进行验证。方法纳入2019年11月至2021年11月于山东中医药大学附属医院住院治疗的437例感染患者,收集患者的临床资料,包括性别、年龄及入院后48 h内的第一次实验室检查指标数据。采用Boruta算法筛选出与AKI发生相关的变量,将筛选出的变量通过XGBoost算法构建AKI临床预测模型,用受试者工作特征曲线下面积(AUC)评估模型的诊断效能,并采用十折交叉法对该模型进行内部验证。结果Boruta算法筛选出12个相关变量:血肌酐、中性粒细胞百分比、淋巴细胞百分比、中性粒细胞绝对值、白细胞绝对计数、尿β2-微球蛋白、血清尿素氮、活化部分凝血活酶时间、丙氨酸转氨酶、天冬氨酸转氨酶、α-羟丁酸脱氢酶、乳酸脱氢酶。由这12个变量构建感染患者AKI的XGBoost临床预测模型,该模型的AUC为0.92,用十折交叉法对该模型进行内部验证的平均AUC为0.77。结论该研究所构建的临床预测模型对感染导致AKI有较好的预测价值,对AKI的早期防治具有指导意义。Objective To construct and validate a clinical prediction model for acute kidney injury(AKI)in patients with infectious diseases.Methods A total of 437 patients diagnosed as infectious diseases admitted to the Affiliated Hospital of Shandong University of Traditional Chinese Medicine from January 2014 to January 2018 were included.The clinical data(including gender,age and indicators of the first laboratory test within 48 h after admission)were collected.The Boruta algorithm was used to screen out the variables associated with the occurrence of AKI,and the screened variables were used to construct a clinical prediction model for AKI by the XGBoost algorithm.The diagnostic efficacy of the model was assessed by the area under the receiver operating characteristic curve(AUC),and internal validation by using ten-fold cross-validation was performed for the model.Results 12 variables were screened out by the Boruta algorithm:serum creatinine,the percentage of neutrophils and lymphocytes,the absolute count of neutrophils and leukocytes,urinaryβ2-microglobulin,blood urea nitrogen,activated partial thromboplastin time,alanine aminotransferase,aspartate aminotransferase,a-hydroxybutyrate dehydrogenase and lactate dehydrogenase.The XGBoost clinical prediction model for AKI in patients with infectious diseases was constructed by these 12 variables.The AUC of the model was 0.92,and the average AUC for internal validation of the model by using ten-fold cross-validation was 0.77.Conclusion This clinical prediction model constructed in this study has a good predictive value for AKI caused by infectious diseases and is instructive for the early prevention and treatment of AKI.

关 键 词：急性肾损伤 感染 XGBoost 预测模型 

分 类 号：R692.9[医药卫生—泌尿科学] R631[医药卫生—外科学]