关键词: 肝脓肿, 肺炎克雷伯菌, 糖尿病, 预测模型, 影响因素

Abstract: Objective To analyze the factors associated with Klebsiella pneumoniae(KP)infection in patients with pyogenic liver abscess(PLA)and to develop and validate an early prediction tool for KP infection. Methods The clinical, laboratory, and imaging data of 287 patients with PLA who were admitted to Qilu Hospital of Shandong University(Qingdao)from December 2016 to October 2025 were retrospectively analyzed. Missing data were handled using multiple imputation by chained equations(MICE). Candidate variables were screened using Lasso regression, and a logistic regression model was subsequently constructed. Model performance was evaluated using receiver operating characteristic(ROC)curves, calibration curves, the Hosmer-Lemeshow test, and 5-fold cross-validation. Sensitivity analyses were also performed. Results Logistic regression analysis showed that a history of diabetes and higher CRP levels were positively associated with KP infection, whereas older age, higher neutrophil(NEU)levels, and a history of abdominal surgery were negatively associated with KP infection. The model achieved an area under the curve(AUC)of 0.837, with a sensitivity of 0.807 and a specificity of 0.767. The directions of association for the key predictors remained stable across different sensitivity analyses. Conclusion History of diabetes, CRP level, age, NEU level, and history of abdominal surgery are associated factors for KP infection in patients with PLA. The predictive model constructed based on these variables shows acceptable discrimination and calibration abilities, and may provide a reference for empirical antimicrobial therapy. However, further validation and optimization using multicenter, larger-sample, and multimodal data are still needed.

Key words: Liver abscess, Klebsiella pneumoniae, Diabetes mellitus, Prediction model, Influencing factors