Found And Modeling Of Risk Factors For Acute Biliary Pancreatitis With Pancreatic Necrosis

Objective:Pancreatic necrosis is the most serious symptom in terms of the local complication of acute pancreatitis(AP),which is difficult to be diagnosed in the early days.If the risk factors of pancreatic necrosis are spotted early and intervened,it will be helpful in the prognosis of pancreatic necrosis.Therefore,the purpose of this research is to investigate the risk factors for pancreatic necrosis in patients with acute biliary pancreatitis(ABP),establishing a predictive model.Methods:The related risk factors of Pancreatic necrosis were explored by derivation cohort,and were verified by validation cohort.In univariate statistical analysis,the?~2test was used for categorical variables,Mann-whitney U test was used for that followed normal distribution in continuous measurement data,.Wilcoxon rank sum test was used for that did not follow normal distribution,and the results were expressed as median(P25,P75)Multifactorial statistical analysis will be adopted logistic regression analysis,being created prediction equations.Eventually,the model performance is analyzed by ROC curve.Results:The incidence of pancreatic necrosis in the derivation cohort is 8.45%(30/355).among which albumin(ALB)(P=0.000)and ascites(P=0.007)are indep-endent risk factors for pancreatic necrosis triggered by ABP.The OR values of them are 0.8505(95%CI:0.7779-0.9299)and 3.3(95%CI:1.3858-7.8592)respectively.The above indicators are finally integrated and constructed to be a predictive model:Z=2.6107-0.1619ALB+1.1940ascites.The area under the ROC curve constructed by the predictive model is 0.8115.The positive and negative predictive values of the model with 0.1 as the cutoff value are 21.59%(19/88)and 95.88%(256/267)respectively,and the negative likelihood ratio is 0.465.The positive and negative predictive values in the validation cohort are 26.2%(17/65)and 99.3%(133/134)respectively,and the negative likelihood ratio is 0.076.The area under the ROC curve is 0.8686.Conclusions:Hypoproteinemia combined with ascites is an independent risk factor for secondary pancreatic necrosis in ABP patients.The predictive model can detect these risk factors effectively,which is a good guide to clinical intervention.