The establishment and validation of a predictive model of early risk factors for acute pancreatitis complicated with infection in ICU

Jiang Jinhua; Huang Peng; Lin Jiandong; Lin Xiao

doi:10.3969/j.issn.1002-1949.2024.01.014

Abstract：Objective To identify the risk factors for predicting the infection in the patients of acute pancreatitis(AP)in the ICU with 24 h after admission,and to construct a predictive model for better decision-making in clinical practice.Methods There were 108 patients with AP who were treated in Intensive Care Unit Department of the First Affiliated Hospital,Fujian Medical University from January 2015 to December 2021 and were collected for modeling,and 155 patients with AP from January 2022 to September 2023 were analyzed for external validation.They were divided into infection group(n =37)and non-infection group(n =71)according to the occurrence of infection.All the patients'clinical characteristics and laboratory examination within 24 h were recorded,and the differences in various indicators between two groups were compared.The independent risk factors for early infection were identified by univariate and multivariate Logistic regression analysis for establishing a risk prediction model of infection.The receiver operating characteristic(ROC)curve was plotted to evaluate the performance of the model,Hosmer-Lemeshow goodness of fit test was used for the calibration evaluation,and the external validation was conducted for stability evaluation.Results In the modeling group,there were 37 patients(34.26%)with infection in total 108 patients.Univariate Logistic regression analysis showed that the patients'age,albumin(ALB),serum creatinine(SCr),procalcitonin(PCT)as well as acute physiology and chronic health evaluation(APACHE)Ⅱscores,Ranson score,bedside index for severity in acute pancreatitis(BISAP),modified CT severity index(MCTSI),the number of organ dysfunction,mechanical ventilation were all the influencing factors for AP patients complicated with infection.Among them,ALB and mechanical ventilation were protective factors rather than risk factors.Multivariate Logistic regression analysis showed an increase in the number of organ dysfunction(OR =1.743,P =0.001)and the elevated SCr(OR =1.005,P =0.027)were associated with concurrent infection in AP patients.Fitted regression model Logit P =-2.267+0.005×SCr(μmol/L)+0.556×the number of organ dysfunction(pcs),the area under the ROC curve(AUC)was 0.768(95%CI 0.672-0.863,P<0.001),with sensitivity of 67.6%and specificity 78.9%,which was better than the other two single factors(P<0.05).The external validation was performed,and the ROC curve showed that the AUC of Logistic model for predicting the risk of infection in AP patients was 0.726(95%CI 0.579-0.874)with some predictive value(P =0.007).Conclusions Increased organ dysfunction number and elevated SCr are early independent risk factors for AP patients complicated with infection in ICU.The predictive model constructed by SCr and the number of organ dysfunction can satisfy various scenarios in clinical practice,and can be persistently used for the prediction of AP infection through the external validation.