Pathogenic distribution,risk factors and prediction model construction of invasive fungal infections in ACLF
HU Weiwei
LIU Liping
ZHAO Renong
SHE Pengyu
LUO Jingyi
Abstract：Objective To investigate the pathogenic distribution of invasive fungal infections(IFD)combined with slow plus acute liver failure(ACLF)and therisk factors and predictive model construction.Methods Patients admitted to our hospital with ACLF secondary to IFD from March 2021 to March 2024 were retrospectively included in the IFD group(n=63),and patients with ACLF who did not develop any infections during the same period were also included in the non-IFD group(n=95).The site of fungal infection and the pathogenic distribution in the IFD group were analysed,risk factors were assessed using logistic regression model,predictive value was assessed using subject's work characteristic curve(ROC),and a decision tree model was constructed.Results Among the 63 patients with ACLF-IFD in this study,the largest partof ca-ses were thepatients with gastrointestinal infections,accounting for 34.92%;63 fungal strains were detected,and Candida albicans was the most frequently isolatedstrain,accounting for 42.86%.Logistic multifactorial results showed that the using duration of the broad-spectrum antimicrobial drugs,invasive manipulation,total bilirubin(TBil),the International Normali-sed Ratio(INR),the end-stage liver model score(MELD)were independent risk factors for ACLF-IFD(P<0.05);albu-min(ALB)and prothrombin activity(PTA)were protective factors(P<0.05).The results of the ROC curve showed that all of the above factors and the combined prediction were statistically significant(P<0.05)for the assessment of ACLF-IFD,with the combination of predictive assessment of ACLF-IFD had an AUC=0.996,95%CI of 0.990～1.000,sensitiv-ity of 0.968,and specificity of 1.000.The decision tree model has a total of 5 layers,17 nodes,and 9 leaf nodes.The model selected 6 risk factors as nodes of the model,MELD,TBil,invasive manipulation,PTA,ALB and INR,of which MELD was the most important predictor,and the classification accuracy of the model was 95.80%.Conclusion The duration of broad-spectrum antimicrobial use,invasive operation,TBil,INR,and MELD are independent risk factors for ACLF-IFD,and ALB and PTA are protective factors for it,so patients with the above mentioned influencing factors need to be paid close attentionduring the treatment period.
Keywords：acute-on-chronic liver failureinvasive fungal infectionstrain distributionrisk factorsprediction model
Publication Date：2025-08-28
Online Publishing Date：2025-09-04(First online date of this platform, not the publication date of the document)
Pages：8( 405-412 )