Development And Validation Of A Risk Model For Predicting In-hospital Mortality In Adult Patients With Infective Endocarditis

Background and Objective:Infective endocarditis(IE)is an infectious disease caused by pathogenic microorganisms on valves and other endocardium,resulting in local inflammation and bacteremia.Despite the advances in diagnosis and treatment,IE still has a high mortality.Therefore,early identification of high-risk patients and adjusting treatments will increase survival rates and improve long-term prognosis.This study aimed to evaluate the role of neutrophil-to-platelet ratio(NPR)to predict in-hospital mortality in patients with IE,then develop and validate a risk model based on NPR.Methods:1.The study retrospectively enrolled consecutive adult in-hospital IE patients from June 2001 to June 2018 in Nanfang hospital as a training cohort.IE patients from the Intensive Care Medical Database(Medical Information Market Intensive Care ?,MIMIC-?)were enrolled as the validation cohort Patients in each set were respectively divided into the survival group and the death group according to hospitalization outcomes.2.In the training cohort,all demographic and clinical data at admission were compared between the survival group and the death group,and the association between variables and in-hospital mortality was assessed by logistic regression analysis.3.Variables with significance in the multivariable logistic regression analysis were included to establish a risk model for in-hospital mortality prediction in IE.The area under the receiver operating characteristic analysis(AUROC)was used to evaluate the discrimination of the model,then the calibration curve and Hosmer-Lemeshow test were used to evaluate the calibration of the model.4.In the training cohort,the Bootstrap re-sampling method was used for internal validation,and the C-statistics was used to evaluate the discriminability of the model.In the validation cohort,external validation was performed with AUROC for discrimination evaluation.Calibration was evaluated by calibration curve and HosmerLemeshow test.Results:1.A total of 294 inpatients with IE in our center were included as the training cohort,with 264 in the survival group and 30 in the dead group;and a total of 312 inpatients with IE were extracted from the MIMIC-III database as the validation cohort,with 258 in the survival group and 54 in the death group.2.In the training cohort,it was found that the admission NPR was significantly higher in the death group than in the survival group(p<0.05).Univariate logistic regression analysis showed that white blood cell count,platelet count,neutrophil to lymphocyte ratio,platelet-to-lymphocyte ratio and NPR at admission,positive blood culture,Staphylococcus aureus infection,double-sided endocarditis,and left-sided infective endocarditis(LSIE)were associated with in-hospital mortality in IE(p<0.05).Multivariate logistic regression analysis showed that the admission NPR(OR=1.097,95%CI 1.046-1.151),positive blood culture(OR=3.789,95%CI 1.236-11.618)and LSIE(OR=7.621,95%CI 1.717-33.814)were independent predictors for in-hospital mortality in IE.The receiver operating curve analysis was performed for NPR.The optimal cut-off value(6.64)had a sensitivity of 0.733 and a specificity of 0.811.The area under the curve(AUC)for NPR was 0.742,with a 95%CI of 0.619-0.865,a positive predictive value of 30.556%and a negative predictive value of 96.396%.3.Based on NPR,positive blood culture and LSIE,a risk model was developed to predict in-hospital mortality in IE and presented as a nomogram with the AUROC of 0.832(95%CI 0.746-0.919).The calibration curve indicated that the predicted probability of in-hospital mortality was consistent with the actual probability(HosmerLemeshow test:p>0.05).4.The C-statistics was 0.820 in internal validation.In external validation,the AUROC was 0.729(95%CI:0.655-0.804)and the calibration curve revealed that the predicted probability of in-hospital mortality was consistent with the actual probability.Conclusions:The admission NPR was an independent predictor of in-hospital mortality in IE patients.The risk model based on NPR,positive blood culture and LSIE had a good performance in in-hospital mortality prediction,which would be applied to clinical practice to early identify high-risk patients.