Association between total bilirubin-to-glucose ratio and long- and short-term prognosis in liver cirrhosis: A study based on the MIMIC-IV database

Critically ill cirrhotic patients generally have poor clinical outcomes, with significantly higher mortality rates compared to non-critically ill cirrhotic patients. Using a large-scale retrospective cohort analysis, this study examines whether the total bilirubin-glucose ratio (TBGR) predicts all-cause mortality in this high-risk patient population. This study assessed the association between the TBGR and mortality risk in critically ill cirrhotic patients using data from the medical information mart for intensive care IV. First, key prognostic variables were selected using the Boruta algorithm, followed by optimal risk stratification with X-tile software. Subsequently, multivariable Cox regression analysis was performed to evaluate the relationship between TBGR and both short- and long-term mortality in critically ill cirrhosis. Survival outcomes were compared using Kaplan-Meier (KM) analysis with log-rank testing, while receiver operating characteristic (ROC) curve analysis assessed TBGR's discriminatory power. Additionally, restricted cubic spline (RCS) regression examined potential nonlinear associations, and prespecified subgroup analyses explored effect modification and Subgroup analyses were conducted to explore potential interaction effects. This retrospective cohort study analyzed 3826 patients from the MIMIC database, with an intensive care unit mortality rate of 26.2%. Using 30-day mortality data from MIMIC, we employed Boruta-based feature importance analysis to validate the prognostic value of the TBGR and determined an optimal cutoff of 1.34 via X-tile analysis (high-risk: TBGR ≥ 1.34; low-risk: TBGR < 1.34). Multivariate Cox regression demonstrated that TBGR was an independent predictor of mortality in cirrhotic patients (P < .001). This finding was further corroborated by KM analysis, which revealed significantly reduced survival in the high-risk group (P < .001). ROC curve analysis indicated moderate predictive accuracy for both short- and long-term mortality. Additionally, the RCS analysis indicated a positive correlation between the TBGR and mortality. Subgroup analyses did not identify significant interaction effects. The TBGR demonstrates a significant association with both short- and long-term mortality in critically ill patients with cirrhosis.