ICU下呼吸道CRKP医院感染风险预测模型及其价值

Risk prediction model for ICU patients with lower respiratory tract CRKP nosocomial infection

目的 探讨重症医学病房(ICU)下呼吸道耐碳青霉烯类肺炎克雷伯菌(CRKP)医院感染患者风险预测模型及其价值。方法 选取2016年1月-2020年12月常州市第一人民医院ICU收治的523例肺炎克雷伯菌(KP)下呼吸道医院感染患者为研究对象,按照6∶4的比例随机分为模型组314例和验证组209例,以受试者工作特征(ROC)曲线下面积(AUC)及Hosmer-Lemeshow检验分别评价模型的区分和校准能力。利用灵敏度、特异度、阳性及阴性预测值判断模型判别能力。结果 模型组Logistic回归显示:入院时已存在其他感染、侵入性操作≥3种、中心静脉插管、急性生理与慢性健康评分Ⅱ(APACHEⅡ)≥20分、使用呼吸机天数≥15 d、使用碳青霉烯类抗菌药物6个变量为CRKP感染的独立危险因素(P<0.05)。其中模型组和验证组的AUC分别为0.864(95%CI:0.825~0.904,P<0.001)和0.832(95%CI:0.782~0.881,P<0.001),H-L检验结果均为P>0.05,提示该风险预测模型具有良好的预测能力。将验证组人群的风险概率值以0.370为界进行模型判别,最终灵敏度为80.95%,特异度为74.66%,阳性预测值为57.95%,阴性预测值为90.08%,诊断准确率高达76.56%。结论 本研究所建立的预测模型判别能力较好,可以为ICU患者是否发生CRKP感染的预警和判别提供参考从而采取针对性的防控措施,规范抗菌药物的使用,降低医院感染率。

OBJECTIVE To explore the risk prediction model for the intensive care unit(ICU) patients with lower respiratory tract carbapenem-resistant Klebsiella pneumoniae(CRKP) nosocomial infection and analyze its value. METHODS A total of 523 patients with lower respiratory tract K.pneumoniae infection who were treated in ICU of the First People’s Hospital of Changzhou from Jan 2016 to Dec 2020 were recruited as the study subjects and randomly divided into the model group with 314 cases and the validation group with 209 cases in a 6∶4 ratio. The capabilities of discrimination and calibration of the model were evaluated by the area under receiver operating characteristic(ROC) curve and Hosmer-Lemeshow test. The sensitivity, specificity, positive predictive value and negative predictive value of the model were observed. RESULTS Logistic regression analysis of the model group showed that 6 variables including other infections at the admission, no less than 3 types of invasive procedures, central venous catheterization, Acute Physiology and Chronic Health Score II(APACHE II) no less than 20 points, use of ventilator no less than 15 days and use of carbapenems were the independent risk factors for the CRKP infection(P<0.05). The AUCs of the model group and the validation group were 0.864(95%CI:0.825-0.904,P<0.001) and 0.832(95%CI:0.782-0.881,P<0.001), respectively. All of the results of H-L tests were P>0.05, indicating that the risk prediction model had good performance in prediction. The risk probability value of the validation group was set as 0370 for discrimination, and the final sensitivity was 80.95%, the specificity 74.66%, the positive predictive value 57.95%, the negative predictive value 90.08%, the accurate rate of diagnosis 76.56%. CONCLUSION The prediction model has good performance in discrimination. It can provide reference for early warning of CRKP infection so as to take targeted prevention measures, reasonably use antibiotics and reduce incidence of nosocomial infection.