Identification of Independent Prognostic Factors and Establishment of Risk Classification Model in Patients with Liver Injury

Abstract

Background: Liver injury is a significant clinical challenge with variable outcomes, highlighting the need for reliable prognostic models for risk stratification. Current scoring systems have limitations in predicting short-term mortality, especially in drug-induced liver injury (DILI) patients. Objectives: To identify independent prognostic factors and establish a risk classification model for predicting 90-day mortality or liver transplantation in patients with liver injury. Methods: This was a retrospective, single-center study conducted from 2020 to 2024, analyzing 223 liver biopsy specimens. Candidate variables included demographic factors, etiological causes, clinical symptoms, serum biomarkers, histopathological grading, and immunohistochemical markers. Multiple imputation addressed missing values. Variables with P < 0.10 in univariate analysis were selected for multivariable modeling using LASSO and stepwise regression. Model performance was assessed with ROC curves and risk stratification. Results: Among 223 patients (70.4% female, mean age 50.3 ± 11.1 years), 92.0% had drug/chemical-induced liver injury. The multivariable model identified five independent predictors: Albumin (ALB; OR = 0.45, 95% CI: 0.31 - 0.67), total bilirubin (TBIL; OR = 1.89, 95% CI: 1.34 - 2.67), aspartate aminotransferase (AST)/alanine aminotransferase (ALT) ratio (OR = 2.15, 95% CI: 1.42 - 3.26), severe lobular inflammation (OR = 3.24, 95% CI: 1.76 - 5.98), and platelet count (PLT; OR = 0.78, 95% CI: 0.65 - 0.94). The model achieved an AUC of 0.818 (95% CI: 0.742 - 0.896). Risk stratification categorized patients into low (score < 20), intermediate (20 - 40), and high (> 40) risk groups with 90 - day mortality rates of 2.1%, 15.7%, and 48.3%, respectively (P < 0.001). Conclusions: We developed and validated a prognostic model incorporating readily available clinical and pathological parameters that effectively stratifies liver injury patients by mortality risk, potentially guiding clinical decision-making and resource allocation.