Risk factors analysis and etiological type prediction of bloodstream infection after liver transplantation
10.3760/cma.j.cn114452-20250531-00318
- VernacularTitle:肝移植后血流感染危险因素分析及病原学类型预测
- Author:
Linting LYU
1
;
Qian HUANG
1
;
Min LI
1
;
Xiaowei MA
1
Author Information
1. 上海交通大学医学院附属仁济医院检验科，上海 200127
- Publication Type:Journal Article
- Keywords:
Liver transplantation;
Bloodstream infections;
Machine learning;
Forecasting model
- From:
Chinese Journal of Laboratory Medicine
2025;48(9):1165-1171
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To identify key risk factors for bloodstream infection (BSI) within 30 days after liver transplantation and to develop an etiological model to predict BSI pathogens based on these factors.Methods:A retrospective study enrolled 122 patients who underwent blood culture after liver transplantation at Renji Hospital, School of Medicine, Shanghai Jiao Tong University from July 2021 to March 2025. Patients were classified into a blood culture positive group ( n=87), comprising non-fermenting Gram-negative bacilli (e.g., Pseudomonas, Acinetobacter, n=8), fungi ( n=8), Staphylococcus ( n=27), Enterobacteriaceae ( n=30), and Enterococcus ( n=14) and a blood culture negative group ( n=35). Baseline variables and laboratory parameters obtained within 24 h of admission and during the postoperative course were subjected to univariate and multivariate analyses to identify risk factors and infection characteristics.Machine learning models for etiological prediction were then developed using these variables. Results:Among the 87 positive cultures, Enterobacterales (30/87, 34.48%) and Staphylococcus (27/87, 31.03%) were the main pathogens. Whithin these two categories of pathogens, Staphylococcus epidermidis and Klebsiella pneumoniae were the primary species, respectively. Autoimmune liver disease was more prevalent in the blood-culture-positive group than in the negative group ( χ2=4.05, P=0.044). The distribution of pathogenic bacteria causing BSI after liver transplantation has certain clinical characteristics. Six-eighths of patients with non-fermenting Gram-negative BSI had underlying malignancies.Among Enterococcal and Enterobacterales BSI cases, viral hepatitis (5/14; 8/30, 26.7%) and autoimmune hepatitis (2/14; 6/30, 20.0%) were more common. The area under curve values of the LightGBM, support vector machine (SVM), and neural network models were all greater than 0.90, indicating that these three models all have high predictive value for the types of pathogens causing bloodstream infections after liver transplantation. Conclusion:The etiological distribution of BSI after liver transplantation exhibits distinct clinical characteristics, LightGBM, SVM and neural network models effectively integrate multi-dimensional data to predict pathogen types, thereby informing infection-prevention and control strategies. Limitations include single-center design and small sample size. Multicenter prospective studies are warranted to validate the model′s efficacy in future research.
