Comparative Evaluation of Gastrointestinal Bleeding Scores for Predicting Decompensation: A Single Center Retrospective Study

Introduction: Gastrointestinal bleeding (GIB) is often associated with hemodynamic instability and decompensation. Despite advancements in medicine, GIB-related mortality rate remains significant. Several scoring systems were developed to predict the risk of decompensation and guide clinical decision-making. However, there is a paucity of data comparing the performance of different GIB scores in predicting decompensation. The objective of our study is to perform a comparison of different GIB scores in predicting important clinical outcomes in a tertiary care center. Methods: We conducted a retrospective study involving 509 patients who presented with GIB symptoms at our tertiary care center between November 2019 and April 2023. We calculated the shock index (SI) and GIB-related scores (ABC, Rockall, and AIMS65). With SI as the reference, receiver-operating characteristic (ROC) curve analysis was performed to compare the predictive potential of the scores for outcomes, such as intensive care unit (ICU) admissions, interventional radiology (IR)-guided embolization, inpatient mortality, and mortality within 30 days. Results: Our study demonstrated that the ABC score exhibited superior performance in predicting 30-day mortality (area under receiver operating characteristic [AUROC] 0.797, P< 0.01) compared to SI (P< 0.01). Similarly, both AIMS65 (AUROC 0.780, P< 0.01) and Rockall (AUROC 0.772, P< 0.01) scores as well showed good predictive potential for 30-day mortality. For the in-hostpital mortality, only Rockall (AUROC:0.84, P< 0.01) and AIMS65 (AUROC:0.78, P = 0.04) scores were superior to SI. In terms of ICU admission, the Rockall score (AUROC 0.76, P< 0.01) showed a strong predictive capability. None of the scores were more predictive of IR embolization (P >0.05). When comparing ABC, Rockall, and AIMS65 to each other, neither showed superiority over the others in predicting any of the four outcomes. However, on subgroup analysis of 255 patients with SI ≥ 1.39, the ABC and Rockall scores showcased superiority for predicting ICU admission (Figure 1). Conclusion: For patients with GIB, the ABC, AIMS65, and Rockall scores demonstrated high accuracy in predicting 30-day mortality compared to SI. For in-hospital mortality, both AIMS65 and Rockall exhibited high predictive performance compared to SI. These findings highlight the utility of these scoring systems in the risk stratification of GIB patients; however, further research is needed to determine the most effective scoring system for predicting specific outcomes (Table 1).Figure 1.: Area under receiver operating characteristic (AUROC) curve showing predicting effectiveness of ABC, AIMS65, and Rockall scores for outcomes: A- 30-day mortality, B- Inpatient Mortality, C- ICU Admission D- Interventional Radiology (IR)-guided embolization. Table 1. - Comparing Various Gastrointestinal (GI) Bleeding-Related Scores for Predicting Outcomes Score AUROC P-Value 30-day mortality ABC 0.7978 0.0044 AIMS65 0.7804 0.0048 Rockall 0.7729 0.0015 Shock Index (reference) 0.6232 - In-hospital mortality ABC 0.7874 0.1139 AIMS65 0.7826 0.0406 Rockall 0.8457 0.0090 Shock Index (reference) 0.6576 - ICU admission ABC 0.7280 0.3306 AIMS65 0.7053 0.5992 Rockall 0.7629 0.0093 Shock Index (reference) 0.6847 - IR-Embolization ABC 0.7856 0.8161 AIMS65 0.8240 0.9307 Rockall 0.8717 0.5053 Shock Index (reference) 0.8153 -