Abstract
BACKGROUND: In recent years, there has been a significant increase in pyogenic liver abscesses (PLAs) caused by multidrug-resistant (MDR) Gram-negative bacteria (GNB), predominantly Klebsiella pneumoniae and Escherichia coli. AIM: To clarify the clinical characteristics and risk factors associated with MDR-GNB-related PLAs, develop a predictive nomogram for personalized risk assessment, and enhance the timeliness of empirical antibiotic selection. METHODS: Based on the antibiotic susceptibility profiles, enrolled patients were divided into two groups: A MDR group comprising 105 individuals and a non-resistant group comprising 163 individuals. A systematic collection of demographic characteristics, laboratory findings, and prognostic indicators was performed. A predictive nomogram was established using multivariate stepwise regression modeling. Model effectiveness was evaluated by examining its discriminative capability, calibration accuracy, and clinical utility through receiver operating characteristic curves with corresponding area under the curve values, calibration graphs, and decision curve analysis. Continuous data were analyzed using the independent-sample t-test if they met normality criteria; otherwise, the Wilcoxon rank-sum test was adopted. For categorical data, Fisher's exact test was chosen when the expected count in any cell was below five; in all other instances, the χ (2) test was applied. RESULTS: This retrospective study analyzed clinical and laboratory data from 268 patients diagnosed with Gram-negative PLA at a major healthcare facility from January 2019 to February 2025. Among these, 105 cases (39%) were associated with MDR-GNB, primarily Klebsiella pneumoniae (43%) and Escherichia coli (42%). Mixed infections were rare, accounting for only 3% of cases. Multivariate regression revealed five independent predictors of MDR-GNB liver abscesses: Age ≥ 60 years, diabetes, presence of a malignant tumor, lower C-reactive protein levels, and prolonged prothrombin time. These variables were integrated into a nomogram to facilitate individualized risk assessment. CONCLUSION: The results imply that being aged over 60, diabetes, malignant tumor, lower C-reactive protein levels, and higher prothrombin time levels can accurately forecast MDR-GNB infections in PLAs, highlighting the importance of early screening to enable more targeted antibiotic treatments. However, as this was a single-center study without external validation, the generalizability of our model remains limited. Future multicenter, multi-ethnic prospective studies are needed to validate and extend these findings.