综合ICU多重耐药菌医院感染的风险模型构建

Construction of risk model for healthcare-associated infection with multidrug-resistant organisms in general intensive care unit

目的：构建重症监护病房（ICU）多重耐药菌医院感染的风险模型。方法对2012年10月—2015年9月入住 ICU＞48 h 的836例患者进行回顾性分析，构建医院感染 logistic 回归模型，对模型进行拟和优度检验、ROC 曲线下面积（AUC）分析。结果3年入住 ICU ＞48 h 的患者共836例，多重耐药菌医院感染发病率为14．23％（119例）。15个单因素分析有意义的自变量纳入 logistic 多 因 素分析，结果显示，ICU 住院时间（OR ＝2．493；95％CI ：1．816～3．494）、基础疾病种类（OR ＝1．536；95％CI ：1．243～1．898）、低蛋白血症（OR ＝87．211；95％CI ：36．165～210．304）、呼吸机插管日数（OR ＝1．723；95％CI ＝1．399～2．121）、发热（OR ＝20．639；95％CI ：3．462～123．043）、原发肺部感染（OR＝0．295；95％CI ：0．133～0．664）变量进入 logistic 回归方程。评价模型效果：灵敏度95％，特异度87．9％，模型 ROC 的 AUC 为0．973。结论 logistic 回归模型对 ICU 患者医院感染预测风险拟合度较好。

Objective To construct the risk model for healthcare-associated infection （HAI）with multidrug-re-sistant organisms（MDROs）in intensive care unit （ICU）.Methods 836 patients who were admitted to ICU for more than 48 hours between October 2012 and September 2015 were analyzed retrospectively,logistic regression model of HAI was constructed,the model was conducted goodness of fit tests and the area under ROC curve analysis. Results Among 836 patients,incidence of HAI with MDROs was 14.23%（n=119）.15 variables that were statis-tically significant in univariate analysis were included in logistic multivariate analysis,the results showed that the following variables entered into logistic regression equation：length of ICU stay （OR,2.493 [95%CI ,1 .816 -3.494]）,underlying diseases （OR,1 .536 [95%CI ,1 .243 - 1 .898 ]）,hypoproteinemia （OR,87.211 [95%CI , 36.165-210.304]）,ventilator days （OR,1 .723 [95%CI ,1 .399-2.121 ]）,fever（OR,20.639 [95%CI ,3.462 -123.043]）,and primary pulmonary infection （OR,0.295 [95%CI ,0.133 -0.664]）.Evaluation of model effect：sensitivity 95%,specificity 87.9%,the area under ROC curve 0.973.Conclusion Logistic regression model has a high goodness of fit in predicting HAI among ICU patients.