Prognostic Risk Factors And Prediction Model Of Acute Pancreatitis Based On EICU Database

Objective:To explore the risk factors for the prognosis of acute pancreatitis,and to build a death prediction model.Methods:The large-scale ICU database eICU database was used to screen out patients with acute pancreatitis.The basic information of patients,laboratory examination results within 24 hours in ICU,past history and complications were collected to predict whether the patients died at discharge.SPSS software was used to analyze the risk factors of death of acute pancreatitis patients,used logistic regression to construct the prediction model,and compared with decision tree model and Apache Ⅳ score to explore the best prediction model.Results:A total of 870 patients were enrolled,including 522 males(60%)and 348 females(40%).72 cases died,8.28%cases died.This study showed that low albumin(OR=0.439,95%CI(0.269,0.716)),low HCO3-(OR=0.909,95%CI(0.853,0.968)),total bilirubin(OR=1.12,95%CI(1.052,1.192)),mechanical ventilation(OR=4.547,95%CI(2.413,8.568)),low map(PR=0.989,95%CI(0.982,0.997)),age(OR=1.057,95%CI(1.034,0.034,respectively),The independent risk factors for death of acute pancreatitis were low SPO2(OR=0.965,95%CI(0.936,0.994)),heart rate(oOR=1.023,95%CI(1.009,1.037)),bun(OR=1.011,95%CI(0.002,1.02)),liver cirrhosis(OR=4.306,95%CI(1.252,14.813)),congestive heart failure(OR=3.518,95%CI(1.202,10.296)).The AUC of logistic regression model,decision tree model and Apache Ⅳ score were 0.885,0.810 and 0.852 respectively,and the Youden index were 0.649,0.538 and 0.6 respectively.The logistic model was simplified,and each factor was assigned according to the or value of each factor in the model.Finally,the acute pancreatitis mortality risk score was made,with the area under the curve of 0.866 and Youden index of 0.583.Conclusion:The logistic regression model is the best in predicting the mortality of acute pancreatitis,with AUC of 0.885 and Youden index of 0.649.The simplified acute pancreatitis mortality risk score can help clinicians quickly and accurately assess the patient’s condition.