住院患者获得碳青霉烯耐药革兰阴性杆菌感染的危险因素分析及列线图的构建

目的：探究住院患者获得碳青霉烯耐药革兰阴性杆菌（carbapenem-resistant organism，CRO）感染的临床危险因素，构建CRO医院感染风险预测模型。方法：查找电子病历选取某院2018年1月-2018年12月CRO及碳青霉烯敏感革兰阴性杆菌（CSO）医院感染的患者为研究对象，按7∶3随机分为训练集和验证集，训练集用于建模，验证集用于模型验证。分析CRO菌株相关信息，采用least absolute shrinkage and selection operator（lasso）回归筛选变量，多变量logistic回归分析构建预测列线图。通过受试者工作特征曲线（receiver operating characteristic，ROC）、校准曲线以及临床决策分析对模型进行内部和外部验证。结果：CRO菌株主要为鲍曼不动杆菌（65.4%）、铜绿假单胞菌（15%）和肺炎克雷伯菌（9.4%），检出以痰液及ICU为主。回归分析结果表明氨基糖苷类、留置尿管、ICU入住时间、抗真菌药物及APACHEⅡ超过20分纳入CRO医院感染预测模型。训练集ROC曲线下面积为0.914（95% CI：0.871~0.956），验证集ROC曲线下面积为0.791（95% CI：0.690~0.891），表明模型具有良好的区分度；校准曲线和Hosmer-Lemeshow检验结果表明模型具有良好的一致性，临床决策曲线分析显示模型具有临床实用性。结论：经住院患者获得CRO医院感染危险因素分析，构建临床风险列线图模型，有助于更好地进行临床碳青霉烯耐药医院感染的防控。

Risk factors and a predictive nomogram for carbapenem-resistant organism infection in hospitalized patients

OBJECTIVE To explore the clinical risk factors of carbapenem-resistant organism (CRO) infection in hospitalized patients and construct a nomogram model for predicting the risk of nosocomial infection.METHODS The patients with nosocomial infection of CRO and carbapenem-sentitive organism(CSO) in the first affiliated hospital of Jiamusi university from January 2018 to December 2018 were included in the study. The data were randomly split into a training set used for modeling and a validation set used for verifying in a ratio of 7∶3. Information about CRO strains was analyzed, variables were screened using least absolute shrinkage and selection operator (lasso) regression, and predictive nomograms were constructed by multivariate logistic regression analysis. Internal and external validation of the nomogram were assessed using the receiver operating characteristic (ROC), calibration plot and decision curve analysis.RESULTS The main strains of CRO were Acinetobacter baumannii (65.4%), Pseudomonas aeruginosa (15%) and Klebsiella pneumoniae (9.4%), and the strains were mainly detected by sputum and ICU. A multivariable model that included aminoglycosides, the urinary tube, length of stay in ICU, antifungal drugs and APACHEⅡ scores over 20 was represented as the nomogram. The model demonstrated good discrimination with the area under the ROC curve analyzed to be 0.914(95%CI: 0.871-0.956) in the training set and 0.791 (95%CI: 0.690-0.891) in the validation set. Calibration plot and Hosmer-Lemeshow test showed good consistency, and decision curve analysis showed that the nomogram was clinically useful.CONCLUSION The analysis of risk factors of nosocomial infection in CRO obtained from hospitalized patients and the construction of clinical risk nomogram model are helpful for better prevention and control of clinical carbapenem-resistant nosocomial infection.