Document resource
Background Sex-based differences in clinical outcomes are increasingly being recognised in acute care, yet few studies have examined whether upper gastrointestinal bleeding (UGIB) risk scores perform differently in male versus female patients. This study evaluated the discriminatory performance of five established UGIB risk scores in predicting inpatient mortality by sex.Methods This retrospective study included 1,003 patients who underwent inpatient oesophagogastroduodenoscopy (OGD) for UGIB between January 2025 and August 2023 at a UK tertiary care centre. Glasgow-Blatchford Score (GBS), CHAMPS, ABC, Japanese score, and MAP(ASH), were calculated for each patient. Binary Logistic regression examined sex as an independent factor for 30-day inpatient mortality, with UGIB aetiology also accounted for (variceal vs non-variceal). Receiver operating characteristic (ROC) curves were used to assess each score’s performance in predicting inpatient mortality separately for male and female subgroups.Results Out of the 1,003 admissions, 59.2% and 40.8% were males and females, respectively, with crude analysis demonstrating female mortality being slightly higher (10.5%) compared to male (8.6%). However, this was found not to be significant (p: 0.304, OR 1.251, CI: 0.816-1.917). These findings were consistent regardless of UGIB aetiology (variceal vs non-variceal).For the 30-day inpatient mortality outcome, the ABC score demonstrated strong discriminative performance in both males (AUC: 0.804) and females (AUC: 0.835). CHAMPS and MAP(ASH) also performed consistently, with CHAMPS yielding AUCs of 0.788 in males and 0.793 in females, and MAP(ASH) showing 0.767 and 0.781 in males and females respectively. GBS performed slightly better in females (AUC: 0.743) compared to males (AUC: 0.691). The Japanese score showed poor discrimination in both sexes, particularly in females (AUC: 0.577). Overall, ABC, CHAMPS, and MAP(ASH) performed reliably across sexes.Conclusion Whilst crude mortality rates were higher in females than males, gender did not show to be an independent predictor of inpatient mortality, which could, however, necessitate a larger cohort size to ascertain validity. Risk scores, particularly the ABC score, maintained high and consistent predictive accuracy for inpatient mortality in both male and female patients with UGIB. These findings support the use of these tools across genders and suggest minimal gender-related performance bias in risk stratification.