Establishment and Validation of a Risk Prediction Model for Sepsis-Associated Liver Injury in ICU Patients: A...

Sepsis-associated liver injury (SALI) leads to increased mortality in sepsis patients, yet no specialized tools exist for early risk assessment. This study aimed to develop and validate a risk prediction model for early identification of SALI before patients meet full diagnostic criteria.

This retrospective study analyzed 415 sepsis patients admitted to ICU from January 2019 to January 2022. Patients with pre-existing liver conditions were excluded. Using LASSO regression and multivariate logistic analysis, we developed a predictive nomogram incorporating clinical variables. Model performance was evaluated through internal validation using bootstrapping method.

Among the cohort, 97 patients (23.4%) developed SALI. The final model identified five key predictors: total bilirubin, ALT, γ-GGT, mechanical ventilation, and kidney failure. The model demonstrated good discrimination (AUC=0.841, 95% CI: 0.795-0.887) and calibration. Decision curve analysis showed clinical utility across a threshold probability range of 4-87%. The model outperformed traditional scoring systems (SOFA and SAPS II) in predicting SALI risk.

This novel nomogram effectively predicts SALI risk in sepsis patients by integrating readily available clinical parameters. While external validation is needed, the model shows promise as a practical tool for early risk stratification, potentially enabling timely interventions in high-risk patients.