Chinese Hepatolgy ›› 2022, Vol. 27 ›› Issue (9): 966-972.

• Frontier, Exploration and Controversy Liver Failure • Previous Articles     Next Articles

Establishment of a prediction model for bacterial infection in patients with HBV related acute-on-chronic liver failure

ZHENG Hui-fang1, LIN Sheng-long2, ZHENG Song1, HUANG Yu-xin1, XIAO Shan-ying1, YE Zi-jie1, LIN Ming-hua2, GAO Hai-bing2   

  1. 1. Fujian Medical University, Fuzhou 350004, China;
    2. Department of Severe Liver Disease, Mengchao Hepatobiliary Hospital of Fujian Medical University, Fuzhou 350025,China
  • Received:2021-11-30 Online:2022-09-30 Published:2022-10-27
  • Contact: GAO Hai-bing,Email: gaohb605@163.com

PDF (PC)

215

Abstract: Objective To investigate the risk factors of bacterial infection in patients with hepatitis B virus related acute-on-chronic liver failure (HBV-ACLF), and to construct a prediction model. Methods A total of 255 patients with HBV-ACLF admitted to our hospital from January 2015 to December 2018 were enrolled, and the clinical data during 2 days after diagnosis were retrospectively analyzed. Bacterial infection occurred during hospitalization was selected as the clinical outcome. The R programming language was used to analyze the data and construct the nomogram. Lasso regression and logistic regression were used to filter variables, analyze risk factors and construct the prediction model. Efficiency of the constructed model was evaluated by receiver operating characteristic (ROC) curve and calibration plot. The model was internally verified by Bootstrap method. Results Among 255 patients with HBV-ACLF, the proportion of male was 78%, and the incidence rate of bacterial infection during hospitalization was 79.60%. Lasso regression analysis Taking lambda = lambda. 1 se (0.049) as the standard, 7 variables including age, direct bilirubin (DBIL), cholinesterase (CHE), prothrombin time (PT), activated partial thromboplastin time (APTT), C-reactive protein (CRP) and hepatic encephalopathy (HE) were selected as the risk factors through Lasso regression analysis. The logistic regression model was logistic(p)＝-7.1733 + 0.0495 × AGE + 0.4107 × ln(DBIL)-0.0002 × CHE + 0.0350 × PT + 0.0610 × APTT + 0.5212 × ln(CRP) + 1.3582 × (HE=1 or 0). Among the 7 variables, AGE (OR=1.05, 95%CI 1.02-1.09), DBIL (OR=1.51, 95%CI 1.06-2.17), CRP (OR=1.68, 95%CI 1.11-2.62) and HE (OR=3.88, 95%CI 1.37-14.10) were independent risk factors. The specificity of the model was 84.62%, the sensitivity was 69.49%. ROC curve showed the new prediction model (AUC = 85.1%) was superior to CRP(AUC = 71.9%) and PCT (AUC = 65.5%). Conclusion In HBV-ACLF patients, older age, baseline hyper DBIL, low CHE level, prolonged PT, prolonged APTT, elevated CRP, and HE are positively correlated with the occurrence of bacterial infection. The model based on these 7 risk factors performs good in predicting the occurrence of bacterial infection during the hospitalization in patients with HBV-ACLF.

Key words: Hepatitis B virus, Liver failure, Bacterial infection, Prediction model