Journal of Clinical Hepatology
Based on a cohort of 490 patients with decompensated liver cirrhosis due to hepatitis C, this article constructs and verifies a machine learning death risk prediction model with random forest as the core. The study finds that ten indicators, including direct bilirubin, cholinesterase, alpha-fetoprotein, prothrombin time, total bilirubin, high-density lipoprotein cholesterol, alkaline phosphatase, immunoglobulin E (IgE), CA19-9, and CA125, have the most predictive value; the AUC of the random forest model reaches 0.811 (reaching 0.926 in some analyses), significantly better than the traditional Child-Pugh (AUC=0.758) and MELD (AUC=0.639) scores, with higher stability and excellent sensitivity under low false positives. The study also reveals that non-traditional indicators such as IgE and CA125 are closely related to inflammation, tumor microenvironment, and metabolic disorders, expanding the biological understanding of liver cirrhosis progression.
This study breaks through the limitations of traditional liver cirrhosis prognosis assessment that depends on a single organ function score, for the first time systematically integrating clinical biochemistry, immunology, and tumor markers. Through an interpretable machine learning model, it achieves dynamic and individualized risk stratification, providing a new tool for precise intervention. Its results help optimize the allocation of medical resources, identify high-risk patients who urgently need liver transplantation or enhanced supportive treatment, and promote the clinical application of new biomarkers such as IgE and CA125 from mechanism exploration. At the same time, it responds to the strategic needs of the national "Action Plan for Eliminating the Harm of Hepatitis C (2021-2030)" to improve the end-of-life management capacity, providing an evidence-based paradigm for the intelligent management of chronic liver disease.
Objective: To construct a predictive model for the risk of 24-month mortality in patients with hepatitis C-related decompensated liver …
