Prognostic Value of the Total Bilirubin-to-Albumin Ratio in Critically Ill Patients with Gastrointestinal Bleeding: A Multicentre Cohort Study.

Abstract

Introduction: The total bilirubin-to-serum albumin ratio (TBAR) has been proposed as an indicator of hepatic dysfunction and systemic inflammation, yet its prognostic relevance in critically ill patients with gastrointestinal bleeding (GIB) remains unclear. This study evaluated the association between TBAR and mortality and developed a TBAR-based prognostic model.

Methods: Data were extracted from the MIMIC-IV v3.1 database, with external validation using the eICU-CRD v2.0. Adult patients with GIB were included after applying predefined eligibility criteria, yielding a final cohort of 1627 individuals. The primary outcomes were 28-day all-cause mortality. Cox regression, Kaplan-Meier analyses, restricted cubic splines (RCS), ROC curves, and subgroup analyses were performed. A prediction nomogram was constructed using variables selected through the Boruta algorithm and evaluated using AUC, C-index, calibration performance, and decision curve analysis.

Results: Nonsurvivors exhibited markedly greater physiological instability, more severe organ dysfunction, and substantially higher TBAR levels. TBAR was independently associated with mortality across all Cox models (fully adjusted HR for 28-day mortality: 1.10, 95% CI 1.06-1.15). High TBAR was associated with more than a twofold increased risk of death. TBAR demonstrated superior discriminatory ability compared with TBIL or ALB alone and provided incremental predictive value when combined with the SOFA score. RCS analysis supported a linear dose-response association. Subgroup analyses showed no significant interactions. The final nomogram-which incorporated SOFA, APTT, TBAR, vasopressor use, CRRT, and anion gap-achieved an AUC of 0.794 and a bootstrap-corrected C-index of 0.789. External validation in the eICU cohort confirmed the robustness of the association and the strong performance of the model (AUC 0.805).

Conclusions: TBAR is an independent predictor of mortality in critically ill patients with GIB. The TBAR-based nomogram demonstrates reliable discrimination and calibration and may facilitate individualized mortality risk stratification in clinical practice.