ICU产超广谱β-内酰胺酶肠杆菌血流感染诊断预测模型的构建和验证OACSTPCD

Objective:Extended-spectrum β-lactamase-producing Enterobacteriaceae(ESBL-E)has become one of the main pathogens in bloodstream infections in intensive care units(ICU),significantly affecting patients prognosis.This study aims to explore the risk factors for ESBL-E bloodstream infections in ICU,develop,and validate a predictive model for diagnosis,providing reference and suggestions for early recognition and control of ESBL-E by clinicians. Methods:A retrospective analysis was conducted on the clinical data of Enterobacteriaceae bloodstream infection cases at Jining First People's Hospital from January 2020 to June 2022.Patients admitted from January 2020 to December 2021 served as a modeling group(n=255),and those from January to June 2022 as a validation group(n=51).Based on ESBL production,patients in the modeling group were divided into a ESBL-producing group(n=131)and a non-ESBL-producing group(n=124).Multifactorial Logistic regression analysis was used to analyze the risk factors for ESBL-E infections,and a diagnostic predictive model was developed.The model's discriminative ability,calibration,and clinical usefulness were assessed using the receiver operator characteristic(ROC)curve,net reclassification index(NRI),calibration plot,C-index,Brier score,decision curve analysis(DCA),and clinical impact curve(CIC).Internal validation was performed using Bootstrap method resampling,and the model was externally validated using data from the validation group. Results:Multifactorial Logistic regression analysis identified the following independent risk factors for ICU-acquired ESBL-E bloodstream infections:acute physiology and chronic health evaluation II(APACHE II)score(>15),nutritional risk screening 2002(NRS 2002)score(≥3),use of cephalosporins,and quinolones within the last 2 months.The regression equation for the ESBL-E infection diagnostic prediction model was-1.718+APACHE II score(>15)×0.989+NRS 2002 score(≥3)×0.989+use of cephalosporins in the past 2 months×0.648+use of quinolones in the last 2 months×0.806.The model showed good predictive ability with an area under the curve(AUC)of 0.831(95%CI 0.781 to 0.881),sensitivity of 79.4%,and specificity of 72.6%.The NRI showed that the model was superior to single-factor models.The Hosmer-lemeshow test showed good fit(P=0.482),internal validation with Bootstrap resampling 1,000 times showed the calibration curve closely approaching the ideal curve,C-index of 0.831,and Brier score of 0.213.The DCA indicated that the model provided a net benefit in the threshold probability range of 0.07 to 0.70.The CIC showed that,above a threshold probability of 0.7,the predicted high risk populations for ESBL-E infection closely matched with the actual occurrence,confirming the high clinical effectiveness of the predictive model.External validation using this model predicted ESBL-E infection in the validation group with an AUC of 0.807,sensitivity of 90.8%,and specificity of 80.0%,showing high consistency with actual outcomes. Conclusion:ICU patients with severe conditions,malnutrition,or recent use of cephalosporins or quinolones are at increased risk of ESBL-E bloodstream infections.The diagnostic prediction model based on[APACHE II score(>15),NRS 2002 score(≥3),recent use of cephalosporins or quinolones]has good predictive value for ESBL-E bloodstream infections in ICU patients.