Abstract
INTRODUCTION: Acute upper gastrointestinal bleeding (UGIB) requires immediate risk stratification in emergency departments (EDs) to optimize resource utilization and patient outcomes. Several validated scoring systems are available for this purpose. The Glasgow-Blatchford Score (GBS) integrates clinical and laboratory parameters to predict the need for urgent intervention or blood transfusion. The pre-endoscopy Rockall score relies solely on clinical variables such as age, hemodynamic status, and comorbidities to estimate risk before endoscopy. The AIMS65 combines five easily obtainable, clinical, and laboratory variables: albumin, international normalized ratio, altered mental status, systolic blood pressure, and age, providing a simple and objective tool for early mortality prediction, which can be rapidly applied in the ED. However, the comparative effectiveness of these tools in predicting diverse outcomes among Malaysian ED patients remains unclear. This study compared the performance of AIMS65, GBS, and pre-endoscopy Rockall scoring systems in predicting clinical interventions and outcomes among patients with acute UGIB presenting to the ED in Malaysia. METHODS: A retrospective cohort study of 293 adult UGIB patients who presented to the ED of a single tertiary academic hospital from January to December 2022 was conducted. Patients were identified using International Classification of Diseases, 10th Revision (ICD-10) coding and were scored using all three systems. Primary outcomes were early blood transfusion (≤24 hours), endoscopy (≤24 hours), ICU admission, rebleeding, and in-hospital mortality. Discriminative performance was assessed using receiver operating characteristic curves with area under the curve (AUC) analysis, and optimal cut-off values were determined. RESULTS: Among 293 patients (median age: 70 years, 60.4% male), GBS demonstrated good performance for predicting early transfusion (AUC: 0.830, 95% CI: 0.782-0.871) and fair performance for ICU admission (AUC: 0.666, 95% CI: 0.609-0.720). AIMS65 showed fair performance for mortality prediction (AUC: 0.717, 95% CI: 0.661-0.768). Pre-endoscopy Rockall demonstrated variable performance (AUC: 0.510-0.667). All systems performed poorly in predicting early endoscopy and rebleeding (AUC: <0.60). CONCLUSION: For transfusion prediction, GBS achieved good performance (AUC: 0.830), while mortality prediction was best achieved using AIMS65 (AUC: 0.717). All scoring systems showed limited utility for predicting endoscopy timing and rebleeding. Score selection should therefore be tailored to specific clinical decisions in emergency UGIB management.