Upper gastrointestinal bleeding (UGIB) remains a common and potentially life-threatening emergency condition requiring early risk stratification to guide clinical management. Although several validated scoring systems such as Rockall, Glasgow-Blatchford Score (GBS), AIMS65, H3B2, ABC, ABL, and Pre-endoscopic Rockall (Pre-RS) are widely used, their discriminative performance for identifying patients at high clinical risk varies across populations.

This retrospective, single-center observational study included 312 adult patients admitted to the emergency department between January 2024 and January 2026 with clinical manifestations of UGIB. Patients were categorized into high-risk and low-risk groups based on clinically significant outcomes, including transfusion requirement, endoscopic/radiological/surgical intervention, intensive care unit admission, rebleeding, or in-hospital mortality.

The primary objective was to develop a novel risk score (HOLD_B), derived from independent predictors identified through multivariable logistic regression analysis. Receiver Operating Characteristic (ROC) curve analysis was used to determine optimal cut-off values for continuous predictors and to evaluate discriminative performance. The newly developed score was compared with established scoring systems using area under the curve (AUC) analysis and DeLong's test for pairwise comparisons.

The study aims to provide a simplified and clinically applicable risk stratification tool for early identification of high-risk UGIB patients in the emergency department setting.

Inclusion Criteria

Age ≥18 years

Consecutive patients presenting to the emergency department between January 2024 and January 2026

Clinical presentation suggestive of upper gastrointestinal bleeding, including hematemesis, melena, or hematochezia suspected to originate from the upper gastrointestinal tract due to rapid intestinal transit

Patients who underwent upper gastrointestinal endoscopy during the index hospitalization

Presence of endoscopic findings confirming upper gastrointestinal pathology

Availability of complete clinical, laboratory, and outcome data in the hospital information management system

Patients evaluated and managed in the Emergency Department of İzmir Atatürk Training and Research Hospital

Exclusion Criteria

Age <18 years

Patients presenting with cardiac arrest at admission

Intubated patients at the time of emergency department presentation

Presence of concomitant active infectious diseases

Patients diagnosed with esophageal variceal bleeding

Use of medications known to increase serum lactate levels, including:

metformin

β2-agonists

methotrexate

zidovudine

linezolid

Patients with missing or incomplete clinical or laboratory data

Patients whose hospital outcomes could not be determined during follow-up

Demographic characteristics, presenting complaints, comorbid conditions, regularly used medications, mental status, vital signs, laboratory values, admission locations, blood transfusion requirements, endoscopic hemostasis procedures, radiological interventions, surgical procedures, length of stay (LOS), readmissions, and in-hospital mortality data were obtained from the electronic patient record system.

All data were recorded by a single physician using a pre-prepared data collection form.