基于MIMIC-Ⅳ数据库的脓毒性心肌病风险预测模型构建与评价  

Construction and Evaluation of a Risk Prediction Model forSeptic Cardiomyopathy Based on MIMIC-Ⅳ Database

作　　者：熊斌[1] 刘银洲 张安[1] XIONG Bin;LIU Yinzhou;ZHANG An(Department of Critical Care Medicine,The Second Affiliated Hospital of Chongqing Medical University,Chongqing 400010,China)

机构地区：[1]重庆医科大学附属第二医院重症医学科,重庆400010

出　　处：《中国医学科学院学报》2024年第5期671-677,共7页Acta Academiae Medicinae Sinicae

基　　金：重庆市卫生健康委医学科研项目(2023WSJK078)。

摘　　要：目的分析脓毒性心肌病(SC)的危险因素,构建与评价SC风险预测模型。方法从MIMIC-Ⅳ数据库中提取脓毒症患者的临床数据,按7∶3比例随机分为训练集和验证集。根据是否患有SC,分为SC组和非SC组。通过单因素和多因素Logistic回归分析筛选SC的独立危险因素,构建风险预测模型,并绘制列线图。采用受试者工作特征曲线下面积(AUC)评估模型的区分度,校正曲线评估模型的校准度,决策曲线分析(DCA)评估模型的临床适用度。结果共纳入2628例脓毒症患者,其中训练集1865例,验证集763例。训练集和验证集SC发病率差异无统计学意义(58.98%比62.25%,P=0.120)。除慢性阻塞性肺疾病(P=0.015)和重症监护室住院时长(P=0.016),其余临床指标两组间差异均无统计学意义(P均>0.05)。Logistic回归分析结果显示,既往患有冠心病(P=0.028)、心力衰竭(P<0.001)、中性粒细胞计数升高(P=0.001)、淋巴细胞计数降低(P=0.036)、肌酸激酶同工酶升高(P<0.001)、尿素氮升高(P=0.042)是SC的独立危险因素。训练集列线图预测模型的AUC为0.759(95%CI=0.732~0.785),验证集列线图预测模型的AUC为0.765(95%CI=0.723~0.807);两个数据集拟合度较好(训练集P=0.075,验证集P=0.067);DCA结果显示列线图预测模型具有良好的临床适用度。结论基于基础疾病及临床生化指标构建的列线图预测模型能够较好地预测SC的发生风险。Objective To analyze the risk factors of septic cardiomyopathy(SC),and to construct and evaluate the risk prediction model of SC.Methods The clinical data of patients with sepsis were extracted from MIMIC-Ⅳdatabase and randomized into a training set and a validation set at a ratio of 7 to 3.According to the presence or absence of SC,the patients were assigned into SC and non-SC groups.The independent risk factors of SC were determined by univariate and multivariate Logistic regression analysis,and a risk prediction model and a nomogram were established.The area under the receiver operating characteristic curve(AUC),calibration curve,and decision curve analysis(DCA)were employed to evaluate the distinguishing degree,calibration,and clinical applicability of the model,respectively.Results A total of 2628 sepsis patients were enrolled in this study,including 1865 patients in the training set and 763 patients in the validation set.There was no significant difference in the incidence of SC between the training set and the validation set(58.98%vs.62.25%,P=0.120).Except chronic obstructive pulmonary disease(P=0.015)and length of stay in intensive care unit(P=0.016),there was no significant difference in other clinical indicators between the two groups(all P>0.05).Logistic regression analysis showed that coronary heart disease(P=0.028),heart failure(P<0.001),increased neutrophil count(P=0.001),decreased lymphocyte count(P=0.036),increased creatine kinase isoenzyme(P<0.001),and increased blood urea nitrogen(P=0.042)were independent risk factors for SC.The AUC of the nomogram prediction model in the training set and validation set was 0.759(95%CI=0.732-0.785)and 0.765(95%CI=0.723-0.807),respectively.The established model showcased good fitting degrees in both data sets(training set:P=0.075;validation set:P=0.067).The DCA results showed that the nomogram prediction model had good clinical applicability.Conclusion The nomogram prediction model based on basic diseases and clinical biochemical indicators can effectively pred

关 键 词：MIMIC-Ⅳ数据库 脓毒症 脓毒性心肌病 预测模型 

分 类 号：R541.9[医药卫生—心血管疾病] R631[医药卫生—内科学]