Document resource
Background Studies in cardiovascular disease have raised concerns regarding racial disparities in the predictive performance of clinical risk scores. However, limited data exists on whether scoring systems for upper gastrointestinal bleeding (UGIB) perform differently in racially diverse populations. This study aimed to compare the performance of five UGIB risk scores in predicting inpatient mortality in White and non-White patients.Methods We performed a retrospective cohort analysis of 1,003 patients who received inpatient oesophagogastroduodenoscopy (OGD) on admission between January 2025 and August 2023 at a UK tertiary centre. The MAP(ASH), Glasgow-Blatchford (GBS), CHAMPS, Japanese scores and ABC scores were calculated using patient’s medical records. Patients subsequently categorised as: Caucasian, African, Asian, and Mixed. We utilized Binary Logistic regression to analyse ethnicity as an independent variable for 30-day inpatient mortality, also further adjusting for UGIB cause (variceal vs non-variceal). ROC analysis using IBM SPSS Statistics v29 was performed separately by ethnicity to assess each score’s predictive performance for inpatient mortality.Results The majority of the cohort belonged to the Caucasian subgroup (91.8%), followed by Asian (1.7%), African (0.8%), and Mixed (1.2%); 4.5% had no racial background recorded. Crude rates showed higher odds of inpatient mortality for the Caucasian subgroup, followed by mixed, Asian, and African subgroups. However, adjusted analysis revealed no significant relationship between race and 30-day inpatient mortality, and these outcomes persisted when adjusted for variceal vs non-variceal bleeding.The ABC score demonstrated strong performance across both racial groups, with improved discrimination in non-White patients. CHAMPS and MAP(ASH) also showed higher AUCs in the non-Caucasian subgroup. GBS and the Japanese score performed better in non-Caucasian patients than in Caucasian patients, though overall discrimination remained lower compared to ABC and CHAMPS. There was no evidence that any score systematically over- or under-estimated risk in the non-Caucasian group.Conclusion Despite crude rates demonstrating different 30-day inpatient mortality rates based on racial subgroup, there was no statistical significance. These findings could, however, be limited by our cohort size and therefore, might necessitate further research.The ABC and CHAMPS scores maintain robust predictive accuracy for inpatient mortality in both White and non-White patients with UGIB. Contrary to concerns raised in other specialities, our findings do not suggest performance bias in these scores by race, supporting their equitable use in diverse patient populations.