Clinical characteristics of pyogenic liver abscess patients with multiple organ dysfunction syndrome and development...

Pyogenic liver abscess (PLA) is one of the most common and severe infectious diseases encountered in clinical practice. It is characterized by rapid progression, and failure to recognize and treat the condition promptly may readily result in septic shock and multiple organ dysfunction syndrome (MODS), thereby posing a serious threat to the patient's life. However, clinical research focusing on the development of MODS in patients with liver abscess remains limited. Therefore, this study aimed to develop and validate a predictive model for MODS in patients with PLA, identify associated risk factors and clinical characteristics, and facilitate the identification of high-risk patients at admission and during the early stage of hospitalization, thereby supporting timely clinical decision-making and management.

A total of 540 patients diagnosed with PLA were retrospectively enrolled for model development. According to the Sepsis-3 criteria, patients were categorized into a MODS group (n = 67) and a non-MODS group (n = 473). Demographic, clinical, laboratory, and imaging data were collected and analyzed. A dual-path feature selection strategy integrating conventional statistics and machine learning was applied, and key variables were identified using the least absolute shrinkage and selection operator (LASSO), the Boruta algorithm, recursive feature elimination (RFE), and the extreme gradient boosting (XGBoost) combined with SHAP (SHapley Additive exPlanations). Independent risk factors were subsequently confirmed using multivariable logistic regression. The resulting predictors showed high consistency across methods and were subsequently used to develop a logistic regression-based prediction model and construct a nomogram. Finally, model performance was evaluated using receiver operating characteristic (ROC) analysis with the area under the curve (AUC), calibration curves, decision curve analysis (DCA), the Hosmer-Lemeshow goodness-of-fit test, and clinical impact curves. To further assess its predictive value, the model was compared with the established quick Sequential Organ Failure Assessment (qSOFA) score. Differences in AUCs were examined using the DeLong test, and DCA was performed to compare clinical net benefit across a range of threshold probabilities. In addition, the net reclassification improvement (NRI) and integrated discrimination improvement (IDI) were calculated to quantify reclassification performance and the incremental predictive value relative to qSOFA.

Among the 540 patients with PLA included in this study, 67 developed MODS during hospitalization. Multivariable logistic regression identified invasive Klebsiella pneumoniae liver abscess syndrome (IKPLAS) (OR = 7.667, 95% CI: 2.397-24.526), septic shock (OR = 4.662, 95% CI: 1.935-11.235), elevated international normalized ratio (INR) (OR = 6.983, 95% CI: 1.274-38.287), and ascites (OR = 13.834, 95% CI: 5.280-36.245) as independent factors associated with MODS. Platelet count was inversely associated with MODS risk (OR = 0.997, 95% CI: 0.994-1.000), suggesting that thrombocytopenia may increase susceptibility to MODS. The predictive model based on these variables demonstrated good discriminative ability, with area under the curve values of 0.826 (95% CI: 0.749-0.903) in the training cohort and 0.749 (95% CI: 0.587-0.911) in the validation cohort. Calibration curves and DCA further supported the model's accuracy and potential clinical utility. In addition, the model outperformed qSOFA in predicting MODS in patients with PLA.