A comparative study of scoring systems that accurately predict the prognosis of lower gastrointestinal bleeding

Abstract

Purpose Lower gastrointestinal bleeding (LGIB) can be fatal; therefore, several scoring systems have been developed to predict its prognosis. We aimed to compare the mortality predictions and evaluate the usefulness of various scoring systems. Methods The medical records of 3,794 patients who visited the emergency department with hematochezia between January 2016 and December 2021 were retrospectively reviewed. We calculated the areas under the receiver operating characteristic curves (AUROCs) for 30-day mortality and prolonged hospital stay (≥ 10 days), based on the age, blood tests, and comorbidities (ABC); AIMS65; Glasgow Blatchford; Oakland; Rockall (pre-endoscopy); and SHA2PE scores. We compared the predictive accuracy of each score. Results Data for 963 patients (median age, 69 years; males, 54.5%; median hospital stay, 6 days) with colonoscopy-confirmed LGIB were analyzed. The 30-day mortality rate was 3.5%. The most common causes of LGIB were ischemic colitis and diverticulum hemorrhage in 19.3% and 19.2% of the cases, respectively. The AIMS65 (AUROC, 0.845) and ABC (AUROC, 0.835) scores were superior in predicting 30-day mortality (both p < 0.001). The SHA2PE score was the most accurate predictor of prolonged hospital stay (AUROC, 0.728; p < 0.001). Through multivariate regression analysis, 30-day mortality was correlated with albumin level ≤ 3.0 g/dL, international normalized ratio > 1.5, blood urea nitrogen level ≥ 30 mg/dL, and systolic blood pressure (SBP) < 100 mmHg. Prolonged hospital stay was correlated with liver cirrhosis, hemoglobin ≤ 10 g/dL, albumin level ≤ 3.0 g/dL, and SBP < 100 mmHg. Conclusion The recently developed scoring systems accurately predict LGIB prognosis, and their usefulness in clinical decision-making was confirmed.