Predicting mortality and readmission in UGIB

This study aimed to assess the accuracy of three pre-endoscopic scoring systems—Glasgow-Blatchford Score (GBS), AIMS65, and pre-endoscopic Rockall Score (pRS)—in predicting 30-day mortality and hospital readmission in patients with upper gastrointestinal bleeding (UGIB) in Iranian tertiary hospitals. Conducted prospectively between April 2024 and April 2025, the study included 290 patients presenting with UGIB symptoms such as hematemesis, melena, syncope, and coffee-ground vomiting. Most patients had severe comorbidities like hepatic failure, malignancy, or heart disease, contributing to a high 30-day mortality rate of 23.4%.

Among the scoring systems, the pre-endoscopic Rockall Score (pRS) showed the highest predictive accuracy for mortality (AUROC 0.815) and readmission (AUROC 0.605). AIMS65 also performed well for mortality prediction (AUROC 0.813) but was less effective for readmission (AUROC 0.548). The Glasgow-Blatchford Score (GBS) demonstrated moderate predictive ability for mortality (AUROC 0.762) and was primarily useful for identifying low-risk patients needing early discharge.

Low-risk thresholds for the scoring systems—pRS < 1, GBS < 2, and AIMS65 < 1—achieved high sensitivity and negative predictive value (NPV), with pRS showing the best balance (sensitivity 95.4%, NPV 87.5%). This highlights its utility in safely identifying patients for early discharge. Laboratory findings such as low hemoglobin and albumin levels and high blood urea nitrogen (BUN) and INR values were strongly associated with mortality, emphasizing the importance of biochemical and hemodynamic parameters.

The study concluded that pRS is the most effective tool for predicting mortality and readmission, supporting its use for risk stratification and resource optimization in emergency settings. However, results may not generalize to non-tertiary care settings, and newer risk models were not evaluated due to resource constraints.