Development Of A Nomogram To Predict The Risk Of Clinically Important Gastrointestinal Bleeding In Patients Receiving Renal Replacement Therapy In The ICU

BackgroundCritically ill patients in the intensive are unit(ICU)are at risk for gastrointestinal bleeding,which is associated with adverse outcomes.Renal replacement therapy is a commonly used organ support technique in the ICU,and it will inevitably disorder the coagulation function of patients and increase the risk of gastrointestinal bleeding.Previous studies and clinical work have observed that the incidence of bleeding in renal replacement therapy patients is higher than that in ordinary patients,and always with the adverse outcomes.Few studies have explored the risk factors for gastrointestinal bleeding in this special population.Taking patients need renal replacement therapy in the ICU as a study population,make accurate prediction of every patients at risk for clinically important gastrointestinal bleeding(CIB)can help doctors and nurses to better choose preventive procedures,and it is necessary to establish a prediction model.ObjectiveThis study aimed to investigate the risk factors for CIB in critically ill patients,and establish an effective prognostic nomogram for Clinically important gastrointestinal bleeding in critically ill patients using renal replacement therapy.MethodsData on patients who underwent renal replacement therapy in the intensive care unit at the Zhujiang Hospital of Southern Medical University between January 1,2015,and January 1,2020,were retrospective collected.For high-dimensional variables,the Least Absolute Shrinkage and Selection Operator(Lasso)regression model was used to select variables,and the multivariable logistic regression was performed to detect the relationship between variables and Clinically important gastrointestinal bleeding.Discrimination and calibration of the predictive model were assessed using the area under curve(AUC)of the receiver operating characteristic curve(ROC)and calibration plot.Internal validation was conducted using bootstrapping validation.Finally,a nomogram was developed as a tool for clinical use to predict the risk of Finally,a nomogram was developed as a tool for clinical use to predict the risk of occurrence.ResultsWe analyzed 34 potential variables of CIB in 730 critically ill patients who using renal replacement therapy.Finally,6 final variables were obtained through lasso regression and multivariate logistic regression to establish the final prediction model.Three independent risk factors for clinically important gastrointestinal bleeding were determined by univariate logistic regression multivariate logistic regression analysis:co-existing chronic liver disease(OR=2.41,95%CI 1.11-4.98),use of parenteral nutrition(parenteral nutrition vs enteral nutrition,OR=1.68,95%CI 1.05-2.66)and shock(OR=2.10,95%CI 1.21-3.72)when entering the ICU.The model displayed good discrimination,with the area under curve(AUC)of the receiver operating characteristic curve(ROC)of 0.70(95%confidence interval:0.7325-0.8371)and good calibration.The C-index value of 0.68 could still be reached in the internal validation.Calibration plot showed a good fitting and calibration.ConclusionWe have developed a prediction nomogram of clinically important gastrointestinal bleeding in critically ill patients who using renal replacement therapy.This novel nomogram might serve as an essential early warning sign of CIB in critically ill patients who using renal replacement therapy.