926 THE RISK STRATIFICATION IN UPPER GASTROINTESTINAL BLEEDING (RUGB) STUDY: AIMS65 IS SUPERIOR TO GLASGOW-BLATCHFORD AND ROCKALL SCORING SYSTEMS IN PREDICTING INPATIENT MORTALITY IN A PROSPECTIVE MULTICENTRE STUDY

Abstract

The American College of Gastroenterology recommends early risk stratification in all patients presenting with upper gastrointestinal bleeding (UGIB). Currently, the most widely used algorithms to predict outcomes in UGIB are the Glasgow-Blatchford (GBS) and Rockall scores. AIMS65 is an alternative risk stratification score validated in retrospective studies to be superior to GBS and Rockall scores in predicting inpatient mortality, however there are few prospective trials. AIMS65 has the advantages of not being weighted and can be calculated upon presentation with routinely obtainable pathology values. To prospectively validate AIMS65 as a predictor of inpatient mortality in patients presenting with UGIB and to compare AIMS65 with GBS, pre-endoscopic (“pre”) Rockall and Rockall scores. Patients presenting with UGIB to 3 Australian centres (2 tertiary hospitals and 1 regional hospital) were prospectively included over a 22-month period from 2016 to 2017. Data was collected by specialist registrars using a purpose-built smartphone app. All patients were risk stratified using AIMS65, GBS, pre-Rockall and Rockall scores. Primary outcome was inpatient mortality. Secondary outcomes were: a composite endpoint of inpatient mortality, in-hospital rebleeding, and endoscopic, radiologic, or surgical intervention; blood transfusion requirement; intensive care unit (ICU) admission; rebleeding; and hospital length of stay (LOS). The area under the receiver-operating characteristic curve (AUROC) was calculated for each score. 570 patients were included in the study. Median age was 68 years (range 18-95) and 64% were male. 372 (65%) patients presented on antiplatelet or anticoagulant therapy and 174 (31%) on a proton pump inhibitor. Overall mortality was 4.7%. AIMS65 was superior to GBS (AUROC 0.89 vs 0.67, p<0.0001) and both pre-Rockall and Rockall scores (AUROC 0.76, p=0.0004 and 0.79, p=0.0059) in predicting inpatient mortality. Mortality and LOS both rose with increasing AIMS65 score and a cut-off score of 3 had maximal sensitivity and specificity to define high and low risk groups (mortality 1.3 % vs. 20.4%). AIMS65 was superior to all other scores in predicting LOS with a Somer’s D score of 0.32 (p<0.001). All scores were equivalent in predicting the need for ICU admission. GBS was superior to AIMS65, pre-Rockall and Rockall (AUROC 0.82 vs 0.69, 0.65 and 0.66 respectively) in predicting need for blood transfusion. AIMS65 was equivalent to GBS and pre-Rockall in predicting the composite outcome (AUROC 0.60, 0.65 and 0.62 respectively) with all inferior to the post-endoscopy Rockall (AUROC 0.73). AIMS65 is a simple risk stratification score for UGIB with superior accuracy to other pre- and post-endoscopy scores in predicting inpatient mortality and LOS in a prospective multi-centre trial.