Acute lower gastrointestinal bleeding (LGIB) accounts for 20% of GI bleeding. Clinical prediction tools exist to identify patients at high risk for severe bleeding. There is limited data comparing these tools and their ability to predict endoscopic intervention, need for blood transfusion, and intensive care unit (ICU) stay. To compare the ability of several clinical scores to predict severe bleeding, endoscopic intervention, and adverse outcomes among patients presenting with LGIB. We performed a prospective observational study of patients admitted with LGIB who underwent colonoscopy at a tertiary-care medical center between April 2016 and September 2017. We obtained demographics, details of hospital course, and adverse outcomes (severe bleeding). Severe bleeding was defined as per prior studies and included: continued bleeding in the first 24 hours of admission and/or recurrent bleeding after 24 hours of clinical stability. Seven risk scores were calculated at admission (Strate, NOBLADS, Sengupta, Oakland, Blatchford, AIMS65, and Charlson Comorbidity Index). Univariable and multivariable logistic regression were used to determine risk of adverse outcomes. Area under the receiver operating characteristic curve (AUC) analysis was used to compare the predictive power of each score. A total of 170 patients underwent colonoscopy for LGIB during the study time frame. Demographics are available in Table 1. Fifty-two percent (n=89) had severe bleeding during their hospitalization. Patients with severe bleeding had lower admission hemoglobins (8.6 g/dL vs 11.1 g/dL; p=0.0001), were more likely to have renal failure (30% vs 17%; p=0.05), require blood transfusion (85% vs 36%; p<0.0001), require ICU stay (49% vs 19%; p<0.0001), and had a longer length of stay (6 days vs 4 days; p=.0009). The best predictor of severe bleeding was the Oakland score (AUC 0.74; p<0.0001), whereas the Strate score performed best for in-hospital re-bleeding (AUC 0.66; p=0.0008) and endoscopic intervention (AUC 0.62; p=0.01). The Blatchford score was best for predicting need for transfusion (AUC 0.87; p<0.0001). In a multivariable logistic regression model adjusting for age, sex, renal disease, malignancy, abdominal examination, admission INR and blood pressure, the strongest individual predictors of severe bleeding were admission hemoglobin (OR 0.78; 95% CI 0.67-0.91; p=0.0015) and albumin (OR 0.39; 95% CI 0.18-0.86; p=0.02). No singular clinical risk tool has the best predictive ability when compared against the others across five adverse outcomes. Of the validated LGIB predictive scores analyzed, the Oakland, Sengupta, and Strate scores were the best in terms of severe bleeding, ICU stay, and re-bleeding and endoscopic intervention respectively. Admission hemoglobin and albumin were the strongest individual predictors of severe bleeding.View Large Image Figure ViewerDownload Hi-res image Download (PPT)