Construction of risk model of liver injury related to parenteral nutrition in patients with severe acute pancreatitis
10.3760/cma.j.cn115667-20230808-00009
- VernacularTitle:重症急性胰腺炎患者肠外营养相关性肝损伤的风险模型构建
- Author:
Gang YUAN
1
;
Xinhong WANG
;
Bo SUN
;
Haiyuan SUN
;
Lina ZHANG
Author Information
1. 海军第971医院消化内科,青岛 266071
- Keywords:
Severe acute pancreatitis;
Parenteral nutrition;
Liver injury;
Risk model
- From:
Chinese Journal of Pancreatology
2024;24(5):364-368
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To construct a risk model of liver injury related to parenteral nutrition (PNALD) in patients with severe acute pancreatitis (SAP).Methods:The clinical data of 176 SAP patients admitted to the 971 Hospital of Navy from January 2019 to August 2021 were retrospectively collected. According to whether PNALD occurred or not, the patients were divided into liver injury group ( n=33) and non-liver injury group ( n=143). Multivariate logistic regression was used to analyze the influencing factors of PNALD in SAP patients. Then decision tree model and multivariate logistic regression model were established based on the screened risk factors. Hosmer and Lemeshow Test calibration curves were used to calibrate the two models, and receiver operating characteristic curve (ROC) was drawn and area under the curve (AUC) was calculated to compare the prediction efficiency of the two models. Results:Drinking history (history of alcohol intake), serum albumin / globulin ratio ≤1.45, prothrombin time (PT)≥18.52 s, PT activity ≤48.96, activated partial thromboplastin time (APTT) ≥45.91 s were all risk factors for PNALD. The ROC curve of the multivariate logistic regression model and the decision tree model was drawn, and calculated AUC of the two models was 0.851 and 0.906, respectively; the sensitivity was 79.6% and 80.8%, respectively; the specificity was 80.5% and 79.6%, respectively; and the Youden index was 0.601 and 0.604, respectively, with good consistency.Conclusions:Low serum albumin/globulin ratio and PT activity, high PT and APTT are all risk factors for PNALD. The PNALD prediction model based on the above risk factors has high specificity and sensitivity.
