Construction of prediction model for risk of Gram-negative bacterial bloodstream infection based on data of patients with blood flow infection from MIMIC-IV database
Objective To construct the prediction model for risk of Gram-negative bacterium(GNB)bloodstream in-fection(BSI)based on data of patients with blood flow infection from the Medical Information Mart for Intensive Care IV(MIMIC-IV)database,aiming to provide a new method for predicting the risk of GNB-BSI.Methods Clinical data and laboratory indexes[blood routine(red blood cells,white blood cells,platelets,etc.),blood biochemistry(potassium ions,calcium ions,chloride ions,bicarbonate,anion gap and urea nitrogen,etc.),coagulation function indicators(INR,PT,PTT)]of 2503 patients with bloodstream infection from MIMIC-IV database were collected.These data were divided into the training set(1752 cases)and the validation set(751 cases)at a ratio of 7:3.The LASSO regression was employed to select the factors influencing the incidence of GNB-BSI in the training dataset.These factors were subjected to multivariate Logistic regression analysis in order to establish a nomogram prediction model for the risk of developing GNB-BSI.The differentiation,consistency,and clinical practicality of the model was assessed using receiver operating charac-teristic(ROC)curve,calibration curve,and decision curve analysis(DCA)in both the training and validation sets.Re-sults Age,cancer,hepatobiliary disease,alcohol abuse,potassium,calcium,bicarbonate,anion gap,and urea nitro-gen were independent factors influencing the risk of GNB-BSI.Logistic regression analysis was conducted to develop a risk prediction model for GNB-BSI(nomogram model)based on these factors.The area under ROC curve of the model was 0.711(95%CI = 0.667-0.756)in the training set and 0.705(95%CI = 0.678-0.733)in the validation set.The calibra-tion curve exhibited satisfactory consistency between predicted and actual outcomes for GNB-BSI(P=0.764).The DCA showed that the nomogram model had good clinical practicability.Conclusion A nomogram model for risk of GNB-BSI was established successfully,which had good predictive performance,and effectively identified high-risk patients.
infection risk prediction modelnomogramGram-negative bacterial infectionbloodstream infectionGram-negative bacterial bloodstream infectionMedical Information Mart for Intensive Care IV database
万方数据
维普
