Abstract: Objective To clarify the distribution characteristics of pathogenic bacteria in patients with pyogenic liver abscesses (PLA) and to construct a risk model for predicting patient prognosis, in order to guide precise clinical diagnosis and treatment. Method A total of 229 PLA patients admitted to the Affiliated Hospital of Chengde Medical University from January 2018 to June 2024 were selected as the training set, and 115 PLA patients admitted from July 2024 to March 2025 were selected as the validation set. Pus samples from liver abscesses were collected for pathogen identification. Patients in both sets were followed for 3 months to assess prognosis. Based on the prognosis in the training set, patients were divided into a poor prognosis group(55 cases) and a good prognosis group(157 cases). General characteristics were compared between the training and validation sets. General characteristics and the distribution of main pathogens were compared between the poor and good prognosis groups within the training set. Multivariate logistic regression analysis was used to identify influencing factors for poor prognosis in PLA and to construct a risk prediction nomogram model. The performance of the model was evaluated using the receiver operating characteristic (ROC) curve and calibration curves. Result In the training set, the composition ratios of Gram-negative bacteria and Gram-positive bacteria among PLA pathogens were 77.36% and 22.64%, respectively. The rates of poor prognosis in the training set and validation set were 25.94% and 25.23%, respectively. There was no statistically significant difference in general characteristics between the training and validation sets (P>0.05). In the training set, the poor prognosis group had older age, larger maximum abscess diameter, higher proportions of patients with comorbid diabetes, comorbid hypoalbuminemia, and Klebsiella pneumoniae infection, as well as higher white blood cell (WBC) count, procalcitonin (PCT) level, and C reactive protein (CRP) level compared to the good prognosis group. All these differences were statistically significant (P<0.05). Multivariate logistic regression analysis showed that increased age (OR=1.451, 95%CI: 1.013-2.080), comorbid diabetes (OR=2.276, 95%CI: 1.512-3.425), comorbid hypoalbuminemia (OR=1.971, 95%CI: 1.196-3.247), high PCT level (OR=2.295, 95%CI: 1.413-3.728), large maximum abscess diameter (OR=2.882, 95%CI: 1.507-5.512), and Klebsiella pneumoniae infection (OR=1.770, 95%CI: 1.155-2.713) were independent risk factors for poor prognosis in PLA (P<0.05). The ROC curve showed that the area under the curve (AUC) for predicting poor prognosis in the training set using the prognostic prediction model was 0.927, with a sensitivity of 81.82% and a specificity of 91.08%. For the validation set, the AUC was 0.913, with a sensitivity of 85.19% and a specificity of 93.75%. The Hosmer-Lemeshow test indicated no statistically significant difference between the predicted probability and the actual probability of poor prognosis for both the training set (χ²=0.174, P=0.676) and the validation set (χ²=0.205, P=0.603) in the calibration curves. Conclusion The main pathogens in PLA patients are Gram-negative bacteria, with Klebsiella pneumoniae and Escherichia coli being the most common. Factors influencing poor prognosis include increased age, comorbid diabetes, comorbid hypoalbuminemia, high PCT level, large maximum abscess diameter, and Klebsiella pneumoniae infection. The risk prediction nomogram model constructed based on these factors demonstrates good clinical performance.

Key words: Pyogenic liver abscess, Pathogen distribution, Prognosis, Risk factors, Prediction model