Validation and Performance of Five Upper Gastrointestinal Bleeding Prognostication Scores in a Single-Centre United Kingdom Cohort of over 1000 Patients

Abstract

Objectives: Upper gastrointestinal bleeding (UGIB) is a common medical emergency associated with substantial morbidity and mortality. Several clinical risk scores are available, but their comparative performance across outcomes and patient subgroups remains uncertain. We compared five established UGIB risk scores for predicting inpatient mortality and complications in a large United Kingdom (UK) cohort and evaluated performance according to bleeding aetiology and sex. We also explored the incremental prognostic value of body mass index (BMI). Methods: We conducted a retrospective cohort study of 1003 hospitalised patients with UGIB who underwent oesophagogastroduodenoscopy at a UK tertiary centre (2023–2025). Five risk scores were evaluated: age, blood tests, and comorbidities (ABC); mental status–anaesthesiologist score–pulse–albumin–systolic blood pressure–haemoglobin (MAP(ASH)), Charlson comorbidity index–hospital-onset–albumin–mental status–Eastern Cooperative Oncology Group performance status–steroids (CHAMPS), Glasgow–Blatchford score (GBS), and the Japanese score. The primary outcome was inpatient mortality, and the secondary outcome was an inpatient complication composite comprising rebleeding, repeat endoscopy, or interventional radiology. Discrimination was assessed using area under the receiver operating characteristic curves (AUROCs), with pairwise comparisons using the DeLong test. BMI was evaluated using multivariable regression and AUROC analysis. Results: ABC demonstrated the highest numerical discrimination for inpatient mortality (AUROC 0.817), followed by CHAMPS (0.791) and MAP(ASH) (0.774), with no statistically significant differences between these three scores. For inpatient complications, MAP(ASH) had the highest numerical AUROC (0.697), although its performance did not differ significantly from CHAMPS, GBS, or ABC. Exploratory subgroup analyses demonstrated generally good mortality discrimination for ABC and CHAMPS across bleeding aetiology and sex. Adding BMI to ABC increased the AUROC from 0.817 to 0.828, although this was not statistically significant (p = 0.224). Conclusions: ABC, CHAMPS, and MAP(ASH) demonstrated the strongest discrimination for inpatient mortality, while discrimination for inpatient complications was more modest across scores. These findings do not establish the clinical superiority of a single scoring system, and BMI did not provide significant incremental prognostic value.